Posts

Predicting Online Customer Intent

When a customer walks into a store, salespeople work to understand what the customer’s purchasing needs are and decide pretty quickly if they are window shoppers, impulse buyers, bargain hunters, researchers, regular customers, or even just wandering around and not shopping at all. How you respond in the physical world is no so different than in the digital world. But how do you respond to customer needs when there is no salesperson in the room? The answer is by using predictive modeling linked to web analytics. 

Understanding the customer segmentation by propensity score 

The analogy of a physical shop translates into eCommerce websites and into predictive analytics. Visitor and customer behaviour patterns can be anonymously observed by your first party web analytics data and using machine learning, you can predict who will buy. A model can score behaviours according to propensity to buy giving a value between 0 and 1 where 1 would be 100% likelihood to buy. Customers with a value closer to 1 know what they want and are heading for till so they will tend to buy without help, so there is no need to waste effort and budget on these customers that would buy anyway.  Customers with lower scores will need much more effort to actually get them closer to any meaningful conversion behaviour, this may not be the best course of action to take given how much resource may be needed to achieve that. The greatest opportunity lies with focussing on the customers who score around 0.5; the floating middle, as their behaviour shows they are much more open to persuasion. Those that can be persuaded and influenced given the right nudge can be targeted and brought to sale. This describes how an predictive propensity scoring model ML model can help for a single conversion goal. The opportunity grows wider with more goals added – such as scoring consumer interest across a range of different product categories or offers, and using that to drive a personalised response. 

Furthermore AI machine learning (ML) can also help understand how different responses to the customer can shift the likelihood of them taking a particular action, or buying a particular product. Understanding customer intent using propensity scores can guide you into offering the most targeted offer or incentive at the right time for the biggest impact. 

Customised user journey and ROI efficiency 

One of the main advantages of using predictive modelling is the ability to approach and engage your customer where and when it is the most efficient in order to move the user further in your online sales funnel. This approach can greatly improve and personalise the user experience by allowing your team to identify the customer intent on a case by case basis. For example you can target a customer who is looking for a particular product category with a special offer to increase the conversion potential; or you could provide help to another customer who is identified as having trouble searching for a particular category or product. Similarly, knowing the customers with the high conversion potential will give your marketing team a tool for creating immediate activation actions, like search retargeting. Having this insight maintains the momentum nudging the customer towards the conversion stage in the decision making process which is efficient and targeted. Just like the in store salesperson. 

Customer journey optimization and attribution

Another valuable outcome of predictive intent modelling is the ability to optimise your marketing channels by analysing the ROI for each feature of the website. These insights are extremely relevant for your UX and developments team and will help you to understand which features and touchpoints to focus on, or which ones should be modified or even removed. The goal should be to develop a data-driven approach to the buying barriers and create a quick and easy sales funnel, which is extremely important for maximising the conversions. Again using the offline analogy, this would be like asking your in store sales team to use their knowledge and insight to help design your store experience so that it works best for the customers.

Dynamic prediction based on user behaviour change

There is a huge advantage to be gained by assigning a changing score to each user as they move around your site and provide you with more clues as to their interests and intent. This approach will allow you to adapt quickly to any behavioural change as the prediction model responds to the new incoming data about your website users. Moreover, various customer characteristics such as location, interests and buying patterns can be identified and linked to your sales and marketing strategy. This helps build a profile of customer preferences for future use in other digital marketing channels such as social media, search or display advertising. 

Ultimately, these well timed nudges with additional marketing messages or promotions, can change behaviours and ultimately stimulate more sales and drive commercial goals. Knowing your customer has always been a crucial part of marketing strategy efficiency and data driven approach will allow business to make smarter and better informed decisions, while building intelligent and long sustained relationships with customers. 

Get in touch with us for expert advice!

Thanks for reading – we hope you learned something through this high-level tour of marketing effectiveness methods. If you want to learn more about expert data analytics using AI, or are interested in how Metageni can help you use your data to grow online, then do get in touch with us: hello@metageni.com

Gabriel Hughes PhD


Does this article resonate with you? Check out other posts to the right of this page.

Can we help unlock the value of your analytics and marketing data?

Metageni is a London UK based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com

Attribution is changing – how can you prepare?

Online marketers have started to understand that with the imminent demise of cookie tracking, Multi Touch Attribution (MTA) will never be the same again. But what exactly is happening and how can you prepare? During this era of rapid e-commerce growth and increasing use of digital marketing, it is more important than ever to understand the technologies affecting marketing measurement and data privacy. Marketing Attribution, which measures marketing to sale touch point by touch point, has implications for both topics, and the changing nature of browser technology and growing concerns over customer privacy, mean that attribution models will need to change. Marketers will need to adapt and plan for a long-term strategy that allows them to leverage first party data instead of relying on external data trackers. 

So, what are the challenges and solutions? Let us make sure we understand the technology shift first, which comes down to a shift away from a long established tracking technology, the browser ‘cookie’.

What are Cookies again? And are they really going away?

If you don’t already know, cookies are a widely used tracking technology that works through small data text files stored on your machine by your internet browser. Once you have accepted the use of cookies on a website domain, their cookies can  track your behavior on that site. Although strictly limited in data they collect, cookies are a powerful tool to track your user journey from point A to point B. You can think of a cookie as labelling each user with data about their observed behaviour and information they share with that site. The power for marketers comes from using this information for personalisation and even predictions about future visits to the site, including data driven attribution predicting whether a visitor will buy something. 

The most visible application of cookies in marketing has been advertisers using cookies to target consumers with numerous ads and promotions depending on their interests,  including re-targeting you after you visited their website  – this happens especially if the potential customer allowed the use of ‘third-party’ cookies. But what is the difference between first-party and third-party cookies?

What is happening to cookies?

Third-party cookies are dead! – Long live the cookie

The key thing to understand is that only certain cookies are affected by the recent privacy motivated shifts, and that certain cookies, specifically, ‘first party’ cookies, are here to stay. First party cookies are created by the website that you are visiting at the current moment and are quite harmless from a privacy perspective. First party cookies are typically used to personalise and improve your user experience. They cannot ‘spy’ on private data about you or read any files on your computer. It is generally accepted that retailers should be allowed to cultivate automated online relationships through interactions with their customers. Indeed, it is crucial to build up trust with consumers who in return are willing to forgive brands utilising this strictly limited and anonymous data to help market their brand to different audiences.

Third-party cookies on the other hand, are created by other domains usually to harvest your data and understand the underlying information and habits, for advertisers to predict your buying and searching behaviours across different web pages. These cookies have allowed ad-tech companies to create detailed profiles of users for building highly targeted marketing strategies. This is possible because third-party cookies not only allow for tracking the user journey on a particular website domain but across multiple domains, for any domain which shares the third party code, For example, if you visit a healthy eating website and accept the third-party cookie, other websites can be given access to that cookie, and you will likely receive a load of health and food related ads wherever you go online. Although this common use is usually fairly harmless, if sometimes annoying, it is the technical potential of the third party cookie to endlessly accumulate detailed information about a person from site to site. This has caused a concern among data protection advocates and privacy conscious consumers.  

In the end, technology which relies on third-party cookies has lost this battle and they are being phased out. Indeed if you are reading this article in 2023 or later, you will be reading about a technology that used to exist. What started as a tracking blocking feature of less well used browsers  like Safari and Firefox in 2013 has now become the norm, notably since Google announced in 2020 they would phase out third party cookies from Chrome which is used by over 50% of users. Google’s new ‘privacy first web’ will change the digital marketing space forever. 

As cookies are being phased out new tracking technologies are growing in importance, such as user cohorting (e.g. Google’s ‘FLOC’) and anonymous IDs (e.g. Unified ID 2.0). These are interesting but we think will be limited in terms of scale and no doubt will attract privacy concerns as well.  

The End of ‘View-Through’ Attribution

The ‘view-through’ metric was never the best way to measure online ads. The idea was simple enough – when someone clicks on an online ad the performance can be measured by counting the clicks that convert for ‘click-through’ conversions, and so why not also count the ad views that lead to conversions as well, and call them ‘view-through’ conversions!? The only problem was how to count ad views when customers never actually interact with the ad before making a purchase. Enter once again the third party cookie, which could be dropped by the ad-server domain through the ad placement itself and then updated again later in the online conversion for that advertiser, in order to count each view-through sale. 

What seemed like a clever way to understand how ad impressions drive value for advertisers, the ‘view-though’ rapidly became a major source of mis-attribution. This is because the sale could happen many days after the user is shown the ad in which time many other ads and influences both offline and online may have occurred. The concept of measured ‘viewability’ helped this a little, but over attribution remained. One of the huge challenges has been targeting itself: a well targeted ad in a high reach campaign is almost bound to generate a ‘view through’ sale since all the ad needs to do is get loaded into a users browser at some point in the days before purchase. In this way display ads, in particular, have been rewarded for just showing up sometime before a sale with the actual incremental impact of the advertising unknown. With a click, the consumer is usually making a conscious choice to visit the advertiser, whereas most ‘views’ just happen because the consumer is online a lot. 

However, as explained, the view-through metric depends on the third-party cookie which is history. View based measurement has to connect a person buying something online to an ad seen earlier on a different website, which means you need the cross domain tracking capability of third-party cookies. 

Just as the third party cookie will not be missed by most consumers, likewise many serious measurement professionals will be glad to see the end of view-throughs. Before they celebrate though, it is worth noting that major sites like Facebook, Google and Amazon, can still link ad views to sales when they handle both the ad delivery and the sale on their platform, which of course they uniquely have the reach and power to do. So there may yet be an afterlife for the view-through metric.  

Multi-Touch Attribution Moves to Click or Visit Only

Third-party cookies made it possible to track both ad impressions and clicks in attribution models, but now it is not possible to collect data in this way. The ad view (impression) touchpoints are dropping out of the picture, and only direct clicks remain within the measurable customer journey. What people do on the web, including what ads they get exposed to, is now hidden by stronger privacy controls. However, an attribution which is based on clicks to the advertiser site can still be picked up by first party cookies and web analytics. Therefore online sellers can only tell if their ads are being noticed when a potential customer shows up by clicking on one.   

Attribution in the future

How Will Marketing Effectiveness Change?

The gap left for online ad view measurement means upper funnel brand activity including display, video and social media all become harder to measure. Performance journeys, such as affiliate and search clicks and including some mid funnel consideration activity like generic search and email, can all continue to be measured using data driven attribution methods. So this type of click attribution must now be combined with other methods to evaluate ad views.

The topic of alternative methods for measuring display and social ad impact is huge in itself, so we will just take a brief tour of the three main options: (1) New measurement solutions offered by the media tech giants (2) AB Tests/ Randomised Control Trials (3) Econometric methods. 

First off, the tech giants are rolling out new measurement systems which promise to track how ad exposure has changed people’s behaviour without the possibility of identifying any particular individuals.  Google has posted an explanation of their approach centred on their Privacy Sandbox, which includes solutions for targeting in a post-cookie world. Google’s FLOC and FLEDGE technologies both get around user ID tracking by aggregating users together in anonymous groups at browser level, with tightly defined criteria for the group definitions which prevent drill down or cross referencing to pick out individuals. FLOC (Federated Cohorts of Learning) groups users by affinity and interest for in-market audience targeting which can then be targeted by advertisers whilst retaining user privacy, while FLEDGE (First Locally-Executed Decision over Groups Experiment) focuses on enabling automated display retargeting at non-user level where interactions with advertisers are stored within the browser and then sold by Google as retargetable segments. Google says it will use these aggregation type methods for anonymous measurement through to conversion. The benefit to users of increased privacy protection also leads to increased advertiser reliance on Google to accurately measure performance with no way for independent measurement or verification.    

A second approach to consider for any kind of ad measurement is to run an experimental AB test, or to be precise, a Randomised Control Trial. This is where a group of people who are not exposed to the ad are compared to a well matched group who are exposed to the ad (control vs treatment groups). Done properly, nothing beats an experiment for accurate measurement of incremental impact. Targeting can help with experiments, helping in ensuring the two groups are well matched and therefore suitable for comparison to make ad impact estimations. The problem now is that targeting is increasingly dependent on the big tech media owners i.e. with Facebook and also with Google FLOC, you can target only within their networks and on their terms. The measurement challenge increases when you want to combine or compare media impacts across more than one media network e.g. a campaign on both Google and Facebook combined. Without 3rd party cookies cross media AB testing is technically very difficult since you cannot be sure if someone in the non-exposed control group on one site might nonetheless be exposed to an ad on another one. Solving this requires some kind of common targeting framework such as matched location geo-targeting (a good option, but not without challenges) or otherwise using an opt-in consumer research panel that can work across multiple networks (requiring research company and media owner support). 

The third type of solution is old measurement tech – Econometrics, or Market Mix Modelling. These are statistical methods which represent the earliest forms of classical machine learning and grew in popularity back in the 1990s as computational power made it possible to estimate a wider variety of more complex models with relatively small amounts of data.  The idea is to use several years of observations of ads and sales, as well as data on economic, seasonal and business drivers, to build a ‘best fit’ model. This model measures how the drivers work together to generate sales, in particular estimating a coefficient (a multiplier) for each media channel to estimate how many sales it generates. The great thing about econometrics is that the data is readily available and can be increased in scope by sampling across multiple locations. Given the fragmentation and complexity of online data, it is not surprising that Econometrics is making something of a comeback. The downside is that it remains a tool of very imprecise high level directional estimation, and requires significant expertise to get right. If attribution is a microscope examining the user marketing journey closeup, econometrics sees it through a telescope in the gloom of a cloudy day. 

The marketing analyst is left with these imperfect tools to try and evaluate the now hidden user journey from ad exposure to sales.  A combination of methods will usually work best – for example using a one off experiment to help validate an econometric estimation.  

What do these changes mean for businesses and how do you adapt?

At the highest levels, the big shift going on here is that big tech companies such as Google and Facebook increasingly keep the data about user interactions with ads within their own walled gardens. This increases the dependency that brands have on these giants especially for targeting and measurement. This suggests the long term strategy for marketers should involve growing their own first party consumer data using sources like web analytics, transactions data, customer data and CRM, rather than relying on third-party data collectors. Likewise, a marketing strategy will have to include mechanisms collecting data from their own data points such as logins, subscriptions, email forms and call centres.

Investing in resources to build brand awareness and capitalising the pull approach of the PPC techniques and keywords targeting will be vital. This allows companies to ‘pull’ potential customers from the web as they are already interested in the product or service as they are browsing.  Advertising partners will continue to be used to reach new audiences and grow the customer base, but ad effectiveness measurement of less tangible brand and impression based impacts will increasingly rely on aggregated data methods which group customers together to protect their privacy. Relying on the media platforms to measure how your customers respond to your campaigns will only increase your dependency on them and provides no independent means of validation on your media spend. 

With the increased focus on first party data it is essential to build a strategy around customer relationship and trust. Since businesses will be focused on their own data collection it will be imperative to build up trust with your consumer, who will, in turn, allow you to use their data. This can be accomplished by following strict data privacy policies and using data for the consumer benefits such as user experience enhancement and content personalisation.

It will also be important to understand the aggregated data, as well as first party individual data, grouping consumers based on their common buying behaviours and search habits. Google’s FLOC and FLEDGE should prove helpful for audience analysis, allowing marketers to choose a target audience, segmented by common buying behaviours and interests. Targeting by the audience and contextual targeting will grow in importance as cookie based targeting falls away. 

In conclusion, utilising a data-driven approach for click based attribution modelling will continue to enable accurate insights for commercial decision making and performance ROI, leveraging first party data. Companies will have to rely heavily on their own data collected web analytics and other customer facing systems in order to understand the true value behind each touchpoint in the customer journey. For data around how customers passively interact – for example, ad exposure effects – aggregated data is the only way forwards. At Metageni we only use anonymised first party data for click attribution and then use econometric approaches to understand the branding potential of display, social as well as all offline media channels. While businesses still have time to adjust to the new reality, the digital marketing space has always been subject to change. The question for brands is are they getting closer to their customers or relying on the tech companies to do it for them? 

Get in touch with us for expert advice!

Thanks for reading – we hope you learned something through this high-level tour of marketing effectiveness methods. If you want to learn more about expert data analytics using AI, or are interested in how Metageni can help you use your data to grow online, then do get in touch with us: hello@metageni.com

Gabriel Hughes PhD


Does this article resonate with you? Check out other posts to the right of this page.

Can we help unlock the value of your analytics and marketing data?

Metageni is a London UK based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com

Marketing effectiveness in the new age of eCommerce

The ability to measure success is key to optimising any business strategy and marketing campaigns are no exception.

Marketing effectiveness = the measure of your marketing plan to optimise efficient sales growth and ROI. This is essential to achieving positive short and long-term results.

Marketing teams typically adopt as many marketing strategies as they can simultaneously hoping to drive leads from one of those to see what works. The reality is that they aren’t then able to say with certainty which strategy is successful and which one is not, since the interactions between these approaches are complex and hard to measure.

This challenge has become greater with the increased rate of business change through digital transformation and disruption, increased competition, and rapid eCommerce growth. This makes it all the more important to understand how customers’ needs are changing and for companies to adapt quickly if they do not want to miss important opportunities.

On top of all this, Covid-19 has now sped up the progression towards increased eCommerce impacting consumer behavior and shifting the balance from offline to online retail. With sudden store closures, many customers were forced to shop online to avoid queuing outside the few shops allowed to open during the lock-down. Globally, brands large and small have shifted efforts into boosting eCommerce with more sales online than in-store for many categories for the first time. Crucially, many believe that new purchasing habits formed during the pandemic will become permanent with a strong long-term shift towards regular online shopping in a wider range of categories.

What is your measurement strategy? 

Considering the digital transformation in the past 20-30 years combined with the post-pandemic challenges, it has never been more important for businesses to adopt a data-driven approach and embrace the challenges of marketing measurement.

Successful companies and forward-thinking leaders have focused their attention on data analytics, understanding the value of knowing more about their customer’s preferences and what they do during the conversion journey in order to increase marketing ROI, using this insight to make informed strategic decisions.

Can we measure marketing effectiveness for customers who are influenced and buy both online and offline? 

With the right capabilities, knowledge, and data-driven approach, the answer to this question is an emphatic yes. In the following, we are going to briefly review some of the approaches companies can adopt in their measurement process, and try to explore the pros and cons of each, without getting overly scientific.

Let’s start with Attribution modelling, a bottom-up digital marketing measurement approach that evaluates the impact of each individual customer touch-point in driving conversations. In other words, a potential customer, before deciding to buy a product, is going to ‘touch’ diverse marketing campaigns on different channels, like a paid search click, a promoted post on social media, an email with a discount, and so on. A multi-touch attribution model assigns credit to different touch-points, in order to understand which strategy is working well and generating sales. This is an amazing source of insight because companies can understand which campaign touchpoints are most highly used and help evaluate which channels work better than others and where in the process they apply. Companies use attribution to adjust future marketing strategies through redistributing their budget to reflect the whole digital marketing funnel. If the attribution model is an accurate reflection of how value is created, then sustained growth and improved marketing return on investment will result.

A key challenge with attribution is that not all attribution models are valid. Simple ‘positional’ models help firms understand the bias in ‘last-click’ sales reporting, but do not give definitive answers. Data-driven models promise more certainty, but even these give results that can vary widely between platforms and vendors. Online retailers should be cautious about black box ‘data-driven’ models which they do not fully understand and do not include indications of objective accuracy.

Nonetheless, multi-touch attribution models (also called MTA) are especially good for evaluating the impact of digital online cross channel campaigns since they provide a more granular and personal level view than traditional aggregate methods like the next approach we will look at: marketing mix modelling.

Market mix modelling (MMM) and econometrics are long-established techniques that use aggregated historical performance data and are useful for high-level analysis of marketing channels and sales drivers. These models estimate the total effect that every marketing channel has on sales while controlling factors like prices and seasonality that impact business performance. For example, if we think about sales happening during the pre-Christmas period, such a model would help you consider the normal seasonal increase in sales as a factor working alongside the marketing plan. 

Market mix modelling provides broader recommendations for how marketers should allocate their budgets to optimize performance and it is especially useful to measure offline channels like TV, Radio, and Out of home. On the other hand, this approach is less precise than attribution and is typically subject to a wider margin of statistical error compared to data-driven attribution, especially when evaluating channels and campaigns at a granular level. 

The key point to understand is that these are ‘top down’ statistical econometric models which estimate relationships between time periods and across geographical areas, rather than the ‘bottom up’ touchpoint-based analysis which can be done with digital marketing data. This is both a strength and a weakness: it is a strength because it can include any sales drivers offline or online, but a weakness because it can easily misestimate if not done correctly, and is only able to reliably pick up larger and more long term relationships.

Finally, marketers need to be aware of the most robust method which is run designed experiments and randomised control trials. These Hold out testing methods are mainly used to check and validate results from other methods since they are most robust for getting a precise measure. The only reason they are not used exclusively for all marketing measurement is that the methods are hard to scale beyond the occasional measurement of 1-2 channels at a time within a single campaign. 

Such experiments are conducted using two matched groups of potential customers: a test group, the audience for whom a test is conducted including a specific marketing variant such as extra spending on a marketing channel; and a control group, an audience group that has the same characteristics of the test group audience but is not exposed to the test marketing activity. Random selection of which people are in which audience group can be used to make sure that the specific marketing variant is the only thing that can explain any change in subsequent purchase behavior, in the most robust type of experiment a Randomised Control Trial (RCT). 

For example, an experiment that wants to test the role of a particular ad allocates specific people to a test group where they are shown the ad being evaluated and to a control group of very similar people where the ads are withheld. Evaluating the difference in the results between the two groups gives a very clear signal as to the net impact of the ad. 

Hold out experiments are the most accurate method of measurement in marketing but they can typically only test one or two things at a time and can be difficult to manage.  Experiments require careful design and do a good job of measuring what specific activity they are designed to measure, but no more. Companies must be prepared to compromise on their marketing plan to get the design to work at a sufficient scale. Despite these challenges, properly designed experiments are the most robust way to measure the effectiveness of specific marketing tactics and so major platforms like Facebook and Google provide tools to allow marketers to test specific marketing tactics in this way and these provide useful benchmarks for other more generalised cross channel measurement approaches. 

Holistic attribution

A complete view of offline and online marketing

It is clear that marketing measurement is critical for determining campaign success, optimizing marketing spend, and driving business growth. Today marketers manage multiple campaigns across diverse media channels and through multiple devices and platforms, so accurate marketing attribution and effectiveness are more important than ever.

The complexity and range of methods represent a huge challenge for marketers and analysts. Companies relying on online and offline channels should consider a combination of both approaches to maximize the advantages of each methodology and to mitigate their limits at the same time. Attribution modelling will assign credit for all sales to digital marketing channels, whereas econometrics will take into account offline channels and non-marketing sales drivers, like price and seasonality. This means they can lead to very different results for the major digital channels, making it very hard for marketers to calculate the true ROI of each channel.

Here at Metageni, we have developed a holistic approach that allows us to combine econometrics and attribution in a single reconciled view of marketing and market drivers. This unified marketing measurement method adopts a data-driven approach that combines the aggregate data obtained from the marketing mix modelling and the person-level data offered by multi-touch attribution into a single comprehensive view.

The aim of this holistic approach is to collect relevant information from all the marketing campaigns taking into account the granular data (MTA) while still considering the broader marketing environment and external factors (MMM). In other words, we get the benefit of granularity and accuracy through digital attribution while also taking full account of the non-digital environment.

Crucial to getting this right is ensuring that attribution is accurate and we achieve this by building predictive attribution models using AI (machine learning) whereby we can say that a model that better predicts sales is more accurate than another. This, by the way, is the same standard that is used to evaluate econometric models. Analytical techniques like the use of ‘hold-out samples and diagnostic metrics help ensure the models are robust.

Using the strengths of both models and combining the different types of data helps to produce consistent ROI results for marketers, allocate marketing budgets with efficiency and drive sustained growth. Great skill and expertise is needed as every business has a unique combination of online and offline factors which drive growth online.

Get in touch with us for expert advice!

Thanks for reading – we hope you learned something through this high-level tour of marketing effectiveness methods. If you want to learn more about expert data analytics using AI, or are interested in how Metageni can help you use your data to grow online, then do get in touch with us: hello@metageni.com

Gabriel Hughes PhD


Does this article resonate with you? Check out other posts to the right of this page.

Can we help unlock the value of your analytics and marketing data?

Metageni is a London UK based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com

Abandon the last click now!
– the guide for attribution newbies

With so much to say about the challenges and methods of attribution and the role of machine learning, we sometimes forget that many folks are still using last-click to measure marketing success and considering improved attribution for the very first time.

Perhaps you are new to digital marketing or analytics and you just need to know what on earth it all means!? If this is you, then here is an introductory guide to attribution – what it is, why it’s important and how you can start to get to grips with it.

What is marketing attribution anyway?

Marketing attribution is a digital version of marketing effectiveness measurement, using the deeper richer data available online. The primary question being asked is what precisely is the value of each different type of marketing activity, in terms of the sales it generates?

Answering this helps you understand what are the best opportunities for increased marketing efficiency and sales growth.

What about ‘last click wins’? What is wrong with it exactly?

If you have read anything on attribution you probably heard mention of ‘last click wins’ and maybe something to suggest you should stop relying on it. If this is all a mystery to you let’s briefly look at what it is and why measurement experts have thrown it into question.

What even the most basic digital analytics allows companies to do is anonymously track a user who visits their website multiple times, and observes whether they make a purchase or not. The company generally knows nothing about these users except their very rough location, what time they access the site and what type of browser and device they are using. All ‘last click wins’ attribution does is link any purchases made to the very last visit the user made to the site before purchasing.

So you might shop around, read up some advice, ask friends on social media, click on one or two different sellers, search again for the first one you liked, and then buy something. The final ‘search again’ action would ‘win’ under the last-click attribution rule, which means that the web and advertising analytics systems would automatically assign sales to that ad channel, for example to ‘branded search click’, and zero value to the earlier visits.

When the company evaluates their digital marketing, the effect of this last-click rule is that their evaluation is locked into this rather big assumption: that the final marketing interaction was the deciding factor in causing you to make that purchase. When you consider that some of these interactions might just be ‘exposures’ to banners or videos, you can start to understand why there might be a problem here. But the biggest problem with the last click wins assumption is that when buying online, especially for larger complex purchases like financial services or holidays, people generally go through a process, or ‘funnel’, from awareness of what is on offer through to consideration, research and then finally a purchase decision.

Marketing is based on the idea that this funnel process can be influenced which calls into question the logic of only evaluating the final action the customer makes before buying? Last-click is too late in the process: it is often a navigational action taken by a customer after they have already made up their mind to purchase.

This ‘last click wins’ attribution logic is still the primary measurement rule used in digital marketing systems. The last click is easy to measure and suits quick low-value purchases driven by performance-only marketing. However, as user behavior and retail continues to move more heavily online, the length and depth of customer interactions before the purchase have become way too complex to measure in this simplistic way.

Reliance on last-click means many companies are undervaluing digital ‘upper funnel’ marketing activity such as generic search, social, video, and email – to name but a few – and overvaluing low funnel channels like brand search, voucher code sites, and retargeting ( – some big generalizations here, in fact, all of these channels can work in different ways so you need to evaluate this yourself).

It follows that if the measurement bias due to last-click could be addressed in some way, then companies could make big improvements in marketing efficiency and sales growth, simply by re-allocating marketing budget away from the perceived high performing channels and into the real marketing drivers of sales.

How did it get to this? And why does it matter?

Marketing effectiveness measurement has been around as long as marketing itself. Imagine yourself in a pre-digital age, trying to measure your marketing. At first, it was easy – just count the number of sales you get when doing some new marketing activity. This is particularly easy to understand for direct marketing activity, like old-fashioned knocking on doors to sell something. For example, you knock on 1000 doors and make 200 sales then your marketing makes 2 sales for every 10 doors knocked on. Add in costs and revenue and this allows you to work out of it is actually worth knocking on any more doors or if it’s better to try a different approach. You can think of this as the real-world analog of last-click wins attribution.

Once a company gets larger it tends to introduce more brand marketing activity – such as posters, TV, or print. This branding activity does not usually lead to immediate sales but instead helps develop a brand reputation and position the brand as having certain features or emotional attributes. The impact of brand marketing is thus generally expected to occur over a longer period of time than more direct activity. What this means for measurement is that future sales, maybe over an extended period of time, could benefit from this brand marketing, and so this should somehow be factored in. Maybe you now sell 20% more, because some of your new customers feel they have ‘heard of you’ before. Getting to this 20% figure is quite hard since the influence of brand marketing will generally occur at the same time as other more direct marketing that carries on day by day.

Another change that takes place as companies grow is that they introduce more marketing channels that increasingly run concurrently. For example, they might introduce overlapping marketing activities such as maybe posters, plus print, plus in-store promotions. This channel overlap makes the measurement task much harder, as you don’t want to ‘attribute’ all your new sales to all of these channels at once, or else you risk over counting the sales and overvaluing your marketing effectiveness.

Eventually, as marketing budgets grow it starts to make sense to ask some clever stats people to make sense of all the marketing spend and sales data. As nineteenth-century Philadelphia retailer John Wanamaker supposedly said: “Half the money I spend on advertising is wasted; the trouble is I don’t know which half”. Before digital came along, the state of the art in marketing measurement was ‘market mix modelling’ which is a technique still used to find statistical relationships between marketing spend on different channels and sales, intending to optimize the marketing mix i.e. the best share of the budget for each channel to get the most sales for your budget (incredibly some companies still charge 1000s to do textbook market mix modelling even to evaluate digital marketing! – more on this later…).

The difference with digital is that we now have this new much more granular data about how different visitors interact and make purchases. Enter digital attribution. The challenge it tries to address is how to use this more granular data to do the same thing as market mix modelling but on a much more granular and detailed level, moving beyond ‘last click wins’.

What is an attribution model?

An ‘attribution model’ is simply the algorithm that assigns sales value to the various marketing channels that might be involved in driving that sale. Remember an ‘algorithm’ can be a very simple rule, and so ‘last click wins’ is actually an attribution model (so in a way, everyone is doing attribution modelling whether they are aware of it or not!).

Beyond this then we have to distinguish between ‘rules-based’ and ‘data-driven’ attribution models. A rules-based model (also known as ‘positional attribution’) is a model which takes the data on a series of interactions and then applies a simple rule to share out the value. Typical models are –

  • • Last-click wins – is problematic, see above
  • • First-click wins – good for new customers, but how far back in time do you go?
  • • Linear – credit shared equally between all touchpoints
  • • U shaped – credit first and last click 40% each, 20% for middle touchpoints
  • • L shaped – credit the first 60% and 40% to all the rest (%’s may vary)
  • • Time decay – more credit the closer the point is in time to the last click

The key benefit of these rules-based models is that they are relatively simple to apply and understand, and so provide a quick and easy way to sense check last-click attribution and see how advertising is working throughout their online journey. The downside is that it can seem like entering a hall of mirrors where there is no longer a true picture of what is going on.

A ‘data-driven’ model, on the other hand, promises to solve the problem once and for all. Data-driven attribution covers a wider range of algorithms, together with the idea that the model has some basis in evidence defining that particular model as superior or more ‘accurate’ than other attribution models.

(It should be stressed that ‘data-driven’, does not necessarily exclude ‘rules-based’ models, as it might be that data analysis is used to define the particular time windows and positional and channel weights applied in a ‘rules-based’ model.)

While data-driven models seem to offer the perfect solution, they have challenges all of their own. There is a huge range of possible ways to get a data-driven model, and rival data scientists and companies all claim that their approach is the best, so there is, in fact, no one true data-driven model. The quality of the data is always worth considering, as ‘garbage in garbage out’, can skew even the most sophisticated methodology. Major questions exist about how to evaluate such models as accurate or not, and what the right criteria for ‘accuracy’ even are for this problem. Newcomers to the topic are advised to be wary of attribution vendors who sweep all these problems under the carpet and insist that their model is somehow the one true model that will solve the attribution problem once and for all.

What are some of the other key attribution issues?

There are lots of attribution topics we have not even covered here. Some major ones that most companies cannot really ignore include –

  • • Attribution for users who switch between devices e.g. research on mobile, then buy on desktop
  • • Online meets offline e.g. where online sales are influenced by TV ads, or where online marketing leads to sales in stores
  • • Managing impression vs. click attribution, including the role of social media
  • • Linking attribution to Customer Lifetime Value, and valuing repeat Vs new customers
  • • Making attribution analysis fast and transparent enough to be tactical and actionable

Each of these topics deserves a long post in itself – for now, be aware that each of these is a complex topic giving rise to many alternatives theories and methods.

How do we get started on attribution?

Understanding the full scope of the challenge means you have already got started. It should be clear from this that there is no simple fix to addressing all the attribution challenges you might face. So a good place to start is to think through and prioritize which attribution issues to tackle first.

To do this, first review the current marketing budgets and how they are allocated. How good is the measured ROI of your largest channels, and could it suffer from attribution bias and error in terms of how value is attributed to each channel?

The quickest way to then challenge your existing assumption is usually not to jump in with a full-blown data-driven modelling project, but rather to look for evidence in the form of simple marketing tests and simple rules-based models which explore alternative hypotheses about how value is created.

It is important to understand that developing improved attribution is part of developing a data-driven marketing approach, which in turn is part of developing a broader data analytics capability. This is a process, not a quick fix. You will need to demonstrate the benefits that you will undoubtedly see in terms of more efficient sales growth from your marketing, and then use that evidence to justify investment in a more robust sustained approach.

There is plenty more online about attribution, but hopefully, if you read this far, you should at least now have a better idea of what it is all about.

Good luck getting to the next level and do get in touch if you think we can help.

Gabriel Hughes PhD


Can we help unlock the value of your analytics and marketing data? Metageni is a London UK-based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com

Evolutionary machine learning meets marketing analytics

In the past couple of years, we have been working with a genuinely game-changing technique, which combines evolutionary computing with machine learning, to produce much more accurate predictive models. This allows us to achieve levels of predictive accuracy using ‘white box’ algorithms, which were previously unobtainable except by ensemble and non-transparent methods. For marketing analytics, this is especially valuable because of the need to be able to interpret how predictions are made, the all important ‘why’ behind data-driven decision making. Marketing Multi-Touch Attribution (MTA) is especially improved by this technique.

In data science terms, technology is about the application of genetic algorithms to both feature selection and parameter selection. Genetic algorithms are an underutilized early technique in machine learning, which are now experiencing something of a revival. The best way to understand the type of challenge they solve is to think of a genetic algorithm as an intelligent search technique that can find the best model from a universe of many trillions of possibilities.

How genetic algorithms help

To explain why this is necessary at all, it is important to understand the scale of the problem. A key challenge facing the data scientist is that before they run any machine learner on training data, they must make choices about what features to use, how to represent those features as predictors in their model, and what hyperparameters to allow for model configuration. The reality is, most models bake in these decisions through somewhat arbitrary assumptions made early in the modeling process.

Good data scientists seek justification for their feature selection and model configuration, in terms of theory and a heuristic understanding of the problem. Far too many just accept the first model that achieves sufficient accuracy after trying only a few alternatives. While there are methods for feature reduction which have a basis in statistics, very few analysts are even aware of the sheer scale of possible alternative models that could be legitimately trained with the very same data source.  

For example, in a dataset with, say, just 50 alternative features, and for a single model requiring only between 3 and 10 predictors selected from these, there are already around 20 trillion different possible combinations that could be modelled. Add in multiple alternative model types, and allow for each model to be configured in different ways, and you have 100s of trillions of possible models. Many problems have a wider range of possible features than this. Finding the best possible model is what data scientists refer to as an ‘NP-hard’ problem, whereby the computational time required must be scaled in proportion to all the possible options. It’s like looking for a needle in the biggest haystack you can imagine.  

Enter the genetic algorithm. While a genetic algorithm does not guarantee you will find the optimal solution, it gives you a great chance of getting close without having to try every possible option. The way it works is by generating a ‘population’ of random solutions and then evaluating each of these against the objective criteria of desirable model properties, such as predictive accuracy, a normal error, and so on. There is a lot that could be said about how to get these objective criteria right, as there are potential pitfalls here – but let’s leave this for another post.

The best models in the population are used to seed a new population in a subsequent ‘generation’ of models, and random ‘mutation’ is also applied to occasionally change these models and keep the population fresh. There are multiple generations in this process, and when we run our genetic algorithms we ensure they spit out each new winner as it appears instead of simply arriving at one final winning solution. The effect is to move towards an optimal solution without merely settling on a ‘local maxima’ i.e. the best solution within a limited range of solutions.

Improved predictive accuracy

The difference between the best guess model and a model derived in this way is simply stunning. Gains of 10-20% in predictive accuracy are not uncommon. For marketing attribution, this means we can produce a model which is validated by the ability to predict on a blind sample, while at the same time being transparent and simple enough to explain how specific marketing interactions work to drive sales. There are numerous other applications, for example at Metageni we use different types of genetic algorithm to help select data samples which match the characteristics of cross-device matched data, to help with that particular attribution challenge.

I guess that we will hear a lot more about this technique in the next few years, as one of several meta analytics processes in the toolkit which help scale and optimize analytics in many domains. We are very keen to hear about the experiences of others using genetic or evolutionary approaches, so please do get in touch if you are working in this area.

Gabriel Hughes PhD


Can we help unlock the value of your analytics and marketing data? Metageni is a London UK-based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com

Why do companies waste millions on marketing?

We keep hearing companies confess to spending huge sums on marketing without really knowing the value of what they are doing. It seems to be that under market pressure it is often safer to push for more sales through tried and tested ‘spray and pray’ marketing approaches than it is to agonize over petty details like, is our ROAS measurement accurate, incremental, and reflecting the contribution of all marketing channels? Here I will explore the broad reasons for this.

Playing it safe

The demands of business growth competitors, seasonality, and company cash flow do not allow the luxury of waiting for perfect measurement, so companies tend to try multiple strategies at once in the hope that if they throw enough marketing budget at the wall some of it will eventually stick. This broad market mix approach actually makes sense for spreading overall marketing budget risk but does so at the expense of optimal ROI.

Yet in all these firms, there is usually at least some attempt at marketing effectiveness measurement. Measurement of ROI / ROAS is very hard to do properly because it requires sound marketing attribution. The gold standard of marketing measurement is to run properly structured experiments with control and treatment groups and statistical confidence measures of uplift. Yet there are so many possible marketing strategies and channels that it is not practical to do this when under market pressure. You cannot test everything at once and often you cannot even afford to ‘go dark’ on even just a part of your market to create that all-important experimental control group. 

If there was only a way, your ad spend would align with incremental sales impacts and you would grow faster with higher market efficiency.

The paradox of cutting ROAS analytics to focus on growth

More scalable approaches to marketing measurement such as marketing effectiveness analytics and attribution are still complex and can be expensive to implement. You need to address issues of data quality and you also need well-managed teams of analysts and data scientists. All this costs money and takes time. So perhaps it’s not surprising that the requirements for sound marketing measurement take a lower priority than the more pressing requirement for sales growth. Paradoxically, measuring success is considering secondary to achieving it when there is so much growth to gain, but this is how many organizations regard investment in marketing measurement and analytics.

Given the complexity, advertisers fall back on softer media-independent measures of ad effectiveness such as surveys of brand awareness and purchase intent. These softer measures provide reassurance that ad money is not completely wasted but does not actually attribute any value or ROI to different marketing channels, so leave major questions unanswered.

Actor fragmentation and lack of team focus

Another major reason why companies waste money on marketing is the fragmentation of marketing teams and agencies. Channels as diverse as TV, search and social require very different strategies and skillsets for sound execution. For larger brands and larger budgets, specialization within marketing channels makes sense, but too often this comes at the cost of poor integration between channels. Once again measurement tends to take a low priority. By their nature, different marketing channels have different systems of measurement and so increase the cost and complexity of getting to an integrated view.

Agency marketing measurement is of mixed quality. It can sometimes be excellent but at heart, agency measurement has a conflict of interest, whereby the agency is essentially marking its own homework. Such bias is rarely malicious or deliberate but more often is a case where there is a systematic tendency to prefer more positive accounts of media effectiveness.

Once again if different agencies are handling your different marketing channels then you cannot expect much insight into how these channels work together. The result is marketing inefficiency through wasted and duplicated spend and also missed opportunities for finding synergy between channels.

The challenge of cross channel measurement is solvable but it requires focus and resources at the level of senior management. The CMO typically needs support from engineering and finance to push through a properly integrated approach. In many organizations, there is a lack of understanding of the nature of the challenge. It’s often not that there isn’t enough data but that there is a lack of expertise around how to leverage that data for improved decision-making in marketing. This means it is doomed to fail since it is not easy to well. Sound multi-channel measurement means different things for different businesses and so often requires a unique approach for each organization. Just as soon as you think you have nailed it, business and market conditions change.

Analyse that data!

Do not let perfect be the enemy of good

One piece of advice we give is just to stick with it and make a gradual transition to a more data-driven approach. It has to be better to make decisions within a rational data-driven framework than to simply burn the marketing budget. Some analysis and experimentation are better than none, and more is better still. Strive to improve and do not let perfect be the enemy of good.

The basic idea could not be simpler: money wasted on an activity that is not working can be better spent leveraging activity that IS working. This increases marketing efficiency and sales growth.

Be less wrong!

Gabriel Hughes PhD


Does this article resonate with you? Check out other posts to the right of this page.

Can we help unlock the value of your analytics and marketing data?

Metageni is a London UK-based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com

Cross-device marketing: the attribution challenge you cannot ignore

Attribution across devices is one of the major measurement challenges for digital advertising particularly affecting how mobile and upper-funnel activity is valued. Many choose to ignore the problem and plump for guesswork. Is this wise? And what are the best strategies for dealing with this key source of measurement bias?

What exactly is the problem here?

The challenge is simply that we cannot measure actual user marketing influences before sales when users switch devices. For small purchases, most people research and buy in a single visit on a single device. But for a larger more complex purchase, such as a holiday, TV, or an insurance policy, often a little research is done on the phone, maybe a bit more on a tablet, and then the purchase might be made on a laptop.   

A website visitor is anonymous unless they explicitly sign in at some point. So to be visible as the ‘same’ person on more than one device, users have to be an existing customer already. Heavily used online services like Amazon, Google and Facebook have done a good job of streamlining user management but the norm for the rest of the internet is sites cannot tell whether visits from different devices are made by the same person or different people. This is a major challenge for advertising measurement since an ad click followed up by a purchase on a different device shows up as the ad as leading nowhere and having no apparent ROI.

This measurement bias generally gets worse with more complex and expensive purchases which involve ‘upper funnel’ channels and have longer multi-touch paths to purchase. The early touchpoints are already penalised by last-click attribution, and then the cross-device challenge penalises them even further.

How big is the issue?

Comscore produces global stats on this using their usage panels and reports that globally around 46% of users use multiple devices within any given month, but the figure is much larger in the more developed market at around 50-60% in countries including the US, Germany and UK. The more people use multiple devices in this way, especially during a purchase process, the greater the issue.

Although the growth in the devices is a relatively recent phenomenon, the tracking issue is not new. For as long as internet use has been the subject of research, analysts have worried about ‘home versus work’ use whereby someone might do some shopping research during a lunch break and then pick it up later at home – in fact, this is still a major measurement issue. Also, note that the cross-device issue is strictly a cross-browser issue i.e. a user who switches browser or app on the same device cannot usually be linked across events either.

The bias in marketing measurement is clear: user journeys appear much shorter and less multi-touch than they really are.

Most companies still overvalue the last click and undervalue earlier touchpoints. With cross-device switching, even a first click positional model will give too much credit to the last click. This is because a single click is both first and last, and for many advertisers, these are the largest single type of user journey visible in their data. In reality, many of these single touch journeys will be from users who have visited before, maybe even very recently, but just on another device.

Multi-touch journeys will also get broken by a switch to a different device. Maybe you make 2-3 clicks doing some research on your phone, then 2-3 more choosing and then purchasing on another device, with a short break in between. Once again the first few touches via upper funnel channels get undervalued and their true contribution to marketing ROI is partially hidden.

How to ‘solve’ cross-device attribution

Even just thinking about taking this issue into account is a major step that the majority of advertisers don’t take.

The first step is to consider how big a problem it is for your business specifically. As with other attribution issues, the more considered and complex purchases, often the higher value ones like holidays, high tech or financial services, tend to have a long path to purchase and therefore a greater attribution challenge. You can use published research to get cross-device estimates but it better is to get some idea of your own customers’ cross-device behaviour, for example looking at the cross report in Google ads reporting which leverages the benefit of their own cross-device tracking.

Many companies have some customer tracking across devices thanks to user login and sign up apps, and email/ CRM – and while this data is heavily skewed to the purchasers (more on this below….) it does at least provide a window into your customer cross-device behaviour which you can use to explore just how big a challenge this is for your attribution.

Since most companies just ignore the issue, educated guesswork may actually be a better than average approach. In an attribution course which I periodically run we do an exercise where participants estimate the size of the cross-device bias, simply by considering the proportion of sales that are likely to be affected by the issue and using this estimate to up weight upper funnel attribution models. Maybe you guess around 40% of your sales involve multi-touch cross-device journeys. This suggests that when comparing first and last click models, the shifts that occur for each channel whether up or down, are weaker than they should be – in reality shifting by a factor of a further 40% or so.

This kind of simple analysis may be enough for you to give an apparently low ROI channel the benefit of the doubt, as your estimates could show the channel driving more upper-funnel activity than initially appears.

Device graphs and other technical solutions

The main thrust of technical solutions fall into either ‘deterministic’, or ‘probabilistic’ methods and are generally a mixture of both. This is not as scientific as it sounds. Deterministic means you can actually track a user using some kind of technology, while probabilistic means you make inferences (guesswork) most robustly and correctly possible given the data you have. Crucially the probabilistic approach depends on some level of deterministic linking since it relies on using information about ‘known’ cross-device behaviour to try to infer the unknown cross-device behaviour.

So the basic idea is to link as many users as you can using a tracking database, and then make a well educated statistical guess as to the rest.

When you consider a cross-device solution you will no doubt encounter platforms that promise that you can leverage their massive ‘device graph’ database to join users up. It sounds like they have all the information you need. A ‘graph’ in this context means a dataset describing a network of relationships, in this case between different devices. The tactic employed by these companies is to draw on data from multiple companies and networks to work map cookies to a wide set of more robust and persistent login based IDs. They, therefore ‘know’ that when there are visits from two different browsers, they actually belong to the same user: this is deterministic linking.

Technology providers also use their linked data to train a model to predict what happens when these ID mappings are not available, and which visits are likely to be associated with other visits from other browsers and locations, and following a similar pattern as observed in the deterministic linking. This modelling process is called probabilistic linking.

Before you decide to shell out the large sums required to use these solutions, there are some really major challenges you need to be aware of.

GDPR and user consent, and other challenges

First, the only way a third party can track someone who is not logged in by mapping IDs based on shared login and tracked users from other sources. This type of large-scale mapping is almost certainly a violation of user privacy and identity under the GDPR legislation, and it is only a matter of time before these platforms have to delete this data. User data which has been properly consented to is likely to be very small relative to the total universe of users out there. After all, would you consent to a company you deal with sharing cross-device usable data with a whole network of other companies?

Second, the data gets out of date quite quickly. Users frequently change logins, and suppliers, and also change and upgrade devices. So a large proportion of users who show up have not been already linked to the device graph ID set, which is only ever a small subset of the total number of users who login in. They cannot track everyone, so, for the most part, they have to estimate.

This brings us to the third issue – and this is actually the biggest one – which is that claims about probabilistic linking are almost certainly overstated. Claims about highly accurate matching rates tend to fudge the difference between precision and recall (look these terms up if you want to understand more). When you dig into the problem, a basic fact hits you square in the eyes: there are many occasions when users do something and then switch devices, and many occasions where users do the same thing, but then do not switch devices. No amount of probabilistic modelling can change this fact.

For example, suppose I see 100 people make 4 clicks on a mobile device, and my data tells me that there is a 3% chance that their next click is on a desktop. This suggests that 3 people should now be modelled as switching to desktop. But which 3 people? There is no way of knowing from the data you have. If you knew, you wouldn’t be using probabilistic linking in the first place! In technical terms, the ‘recall’ is fine, but ‘precision is very weak.

DIY linking

Deterministic linking is clearly superior since this is where we simply have all the data we need to match user to user on different devices. What solution providers do then is to effectively offer deterministic linking based on their database, with probabilistic linking used as a fallback option to ‘plug in the gaps left by this method.

Again, if you are piggybacking on someone else’s user data, then there are major privacy concerns to consider. However, it is worth noting that most companies already have some data allowing them to join up users across devices – what we might call ‘Do It Yourself’ (DIY) linking. For example, if your website asks users to log in each time, and collects an email address, then if they read their email from you on a phone, you can potentially match them from device to device. It seems there should be a way to leverage that.

Of course, the challenge is that this is always going to be a limited percentage of users, representing a big gap in your knowledge of cross-device usage. However deterministic linking is never 100% complete. So one way to leverage it is to try your own probabilistic matching, using the data you can match to make inferences about the data you cannot match.

If you do go down this route you should be aware of another major challenge with this partially linked data, which is that it’s heavily biased towards fully signed up customers and against non-customers. It would be very easy to use your own sign-in data to conclude that people who buy from you have complex cross-device usage, whereas people who do not buy from you have simpler single device interactions. The problem with this is that the people who sign up and thus become trackable tend to be the buyers, and so you inevitably have more visibility on their complex user journey data than for the non-buyers. This kind of data bias can easily generate misleading conclusions.   

An alternative solution

Here is where you have to forgive me for plugging our own Metageni solution which we call a ‘cross-device inference model’. We are interested to know the feedback from the community, so let me explain the principles.

The basic idea is that the known or ‘matched’ cross-device data can be viewed in both its matched and non matched forms, and we can observe how the process of matching itself changes the data distribution. The distributions of interest are relevant data features such as the user path length, device use frequency, and device position in the user path. We use this information to resample from our raw data, to create a sub-sample of non matched data that has a similar distribution of these features.

Thus, unlike probabilistic matching, we give up on the attempt to somehow create new cross-device data out of the unknown and instead settle for trying to make the overall sample more representative of what the complete cross-device data set would look like.

For example, we might find that when known data is linked, there tends to be a reduction in single clicks from tablet devices, as these get matched to multiple clicks on other devices. Let’s say these types of interaction fall by 10%. So before we use the raw data which has no matching ID, we create a subsample that randomly drops around 10% of these single click tablet interactions. We do this but for many different features of the data which we know shift around when the data can be matched. We end up with a data set that is not fully linked across devices, but which includes only non-linked data which is considered representative.

We would love to hear what other experts think about this less aggressive and more GDPR compliant approach.

Cross-device use continues to evolve but is not going away

Whatever you decide to do, hopefully, you can now see that it is best not to ignore the problem. Recent research suggests we may be past the peak of cross-device usage for some users who now just use smartphones for almost all their internet use. Mobile has become dominant over desktop, vindicating those companies that adopted a mobile-first strategy a few years back.  

However cross-device use is in many ways a symptom of a continuing rise in multi-task activity, as we sit listening to music on Spotify, watching a show on YouTube on the TV and playing with holiday ideas on our mobile phone, often all at the same time. Measuring how users move through their digital universe continues to be vital to understanding their behaviour. Marketing analysts ignore this problem at their peril.  

Gabriel Hughes PhD


Can we help unlock the value of your analytics and marketing data? Metageni is a London UK based marketing analytics and optimisation company offering custom solutions for in-house capabilities.

Please email us at hello@metageni.com

What is multi-touch attribution good for?

One of my proudest achievements while working at Google was the award in 2011 of a patent for Multi-Touch Attribution (MTA). Since 2007 I led a small team of data scientists and engineers working on the exposure to conversion journey using Doubleclick data. The solution we eventually patented was the culmination of months of research and client analytics support and formed the basis of the Attribution Modelling Tool which is still the most advanced available feature in Google Analytics.

Working in the field of marketing analytics and attribution today I am struck by how many people still fundamentally misunderstand the purpose of rules-based attribution modelling as we had originally conceived it back then. So let me put the record straight.

The purpose of rules-based attribution is not to provide you with the best alternative to last-click wins attribution. Indeed as many have observed the problem that MTA highlights are that there are many different ways to understand and measure the contribution of different marketing touchpoints in the user journey to sale. This means it is not immediately obvious what the relative contribution of different marketing channels could be.

It has always been clear that a definitive answer to that question lay in the direction of some kind of data-driven or statistical approach. Indeed this is what my current company attempts to do using machine learning and what Google and others attempt to do with sophisticated algorithmic models. So what is Multi-Touch Attribution good for?

Challenging hidden assumptions

Any alternative rules-based model highlights the extent of potential bias in last-click attribution. The background to the development of MTA was the deeply ingrained adherence to the flawed last click model of marketing value. Indeed even today over a decade after rules-based attribution was invented a surprisingly large number of major brands still rely on a last-click view of the world. Last-click is baked into most advertising reporting systems and as a result, companies continue to under-invest in upper funnel sales growth opportunities and fail to realise marketing efficiencies.

Companies use last-click attribution without even knowing about the implicit assumptions that are embedded within this logic. When I give talks to larger groups of agencies and advertisers I often open by asking for a show of hands from all those who are, ‘doing attribution. Naturally, it’s a trick question. If you are not ‘doing attribution’, then you are almost certainly relying on last-click wins attribution without realising it.

Indeed, if you make any claim at all about the sales impact of your advertising, then you are ‘doing attribution’, and so the question advertisers need to ask themselves is how they can do a better job of challenging and testing their attribution assumptions.

Around a decade ago Google and others started to address this huge hidden measurement bias through the introduction of metrics such as the assisted click count. So for example in AdWords ‘Search Funnels’ it became possible to see how brand last-click sales are influenced by generic search campaign, ‘assist clicks’. This was a major step forward but was not a game-changer. Just as the magician’s assistant is a secondary presence on the stage compared to the main performer, then in a similar way the assist click is a secondary player to last-click attributed sales.

This is where alternative models come in: they start with equivalent status to the last click. Using an alternative attribution rule such as, say, first click wins demonstrates the equivalent status of anyone attribution rule to another. The point is that without experimental methods or statistical analysis, one attribution rule is no better than any other.

Indeed, last-click is quite a poor choice compared to other positional models. The specific problem with last-click is that it maximizes the risk of attributing a sale too late within a marketing process which influences the user in a series of steps. If the marketing has any value at all it is there to help a brand convince and influence a person to purchase from them in preference to spending their money elsewhere. Somewhere within this multi-step process, the consumer makes a choice. By the time customers make the last click, many of them have already made up their mind beforehand. Thus many last click actions are purely the final navigational steps in a considerable research and purchase process.

So, alternative attribution models help us expose the flaws in last-click wins attribution.

Is it worth investing in accurate attribution? – Attribution gap analysis

Ultimately we all want to know the ‘true’ attribution picture: what is each marketing touch point and channel actually contributing to sales? – if we know this we can allocate budget for maximum efficiency and sales growth. Getting to that accurate picture is a complex challenge however – at Metageni we advocate a custom data driven model incorporating customer signals and trained using machine learning on your own data. We are often asked whether it worth all that effort? One way to answer is to actually measure the uncertainty itself, but comparing a range of different attribution models and measuring the difference, the ‘gap’, between each model. Attribution gap analysis does this, measuring the range of revenue which is subject to ambiguous attribution. If your customers generally use just one marketing channel and only click once or twice on those ads before purchase, your attribution gap may be quite small. In most cases when we measure the gap between MTA models, we expose a large variation between the value of key channels especially in the upper funnel of the customer journey.

So one thing rules-based models are good for then, is exploring the size of attribution problem and measuring the range of uncertainty. This help CMOs and finance teams work out how much to invest in more accurate attribution, addressing how much uncertain attributed revenue and potentially wasted marketing budget there is in any given year.

Understanding the customer journey

Alternative rules-based models do much more than just move us beyond the last click. They also help users understand the complete measured journey to sale. If you compare a first click model to a last-click model you are directly able to see the patterns of where users start their journey compared to where they end their journey. If you introduce a linear model you can understand all the touchpoints that had an influence along the way. A time decay model helps you to understand the closing stages of the process without early focusing on just the final stage. And so on. In other words, rules-based models are an excellent means of summarising the data on multiple complex paths to purchase online.

The need to be able to summarise this data is clear as soon as you try to look at the raw data about each user journey. Many paths are incredibly complex and unique. As an analyst, you need a way to summarise this information so you can understand at a macro level what the most important patterns are. Rules-based attribution models give you precisely this.

One key feature of the attribution modelling tool in Google Analytics is that it allows the user to compare models one against the other. This is because the aim is not to encourage the user to pick a single alternative to last click wins but to compare multiple models for insight into how marketing works throughout the user journey. This feature was central to the original attribution analytics prototype that we created.

Measuring ROI

So can rules-based attribution tell us anything about the actual value of marketing? Of course, it can. Think of each rules-based attribution model as a hypothesis about how value is created and then you can check this hypothesis against the other hypotheses and against your marketing strategy.

For example, if all models agree that a particular marketing channel drives positive ROI then it looks like a pretty safe bet for ongoing investment. If all models show negative ROI then maybe you should look for a cost-saving. If only the upper funnel models show positive ROI for a channel then consider whether your marketing strategy is to use this channel to drive the early stages of purchase such as research, for example, a review based affiliate.

Put simply rules-based models allow you to consider whether your digital marketing is really working in the way that it is supposed to work.

Ultimately decision-makers want simple answers. Multi-Touch Attribution was an attempt to simplify what is, in fact, an extremely complex analytical problem. This is proved to be both the strength and the weakness of MTA as marketers continue to struggle with knowing exactly what to do with these models.

Data-driven models still involve hard choices

There was never any doubt that a robust data-driven approach or experimental framework would be preferable to multiple rules-based MTA models. The promise of an approach that could yield a single definitive answer is too tempting to ignore. Yet even here the analyst has a duty to ask questions of such models and try to understand the underlying patterns in the data.

One major challenge with data-driven attribution models is that navigational actions like brand search clicks are a good predictor of sales even though they are not a reflection of marketing cause and effect. The issue is essential that marketing influences sales through a psychological mechanism that is not directly observable in behavioural data.

Furthermore, experienced data scientists know that there are multiple data-driven models that can be obtained leading to different results and that a simple objective criterion for selecting between them is not easily available. So, even if you can get to decent data-driven models, it is not nearly so easy as you might imagine arriving at a single unambiguous solution.

We should always be open to multiple interpretations of the data and try to avoid oversimplification. So even if you believe you have the perfect data-driven model I would urge any marketer to use the rules-based models as a benchmark and guide to interpretation.

As Einstein said, ‘Everything should be made as simple as possible, but not simpler.’

Gabriel Hughes PhD


Can we help unlock the value of your analytics and marketing data? Metageni is a London UK based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com

Build your own attribution model with machine learning

Sounds too good to be true? Maybe so, but as machine learning and cloud data technology become more accessible and scalable, building your data-driven Multi-Touch Attribution (MTA) model is becoming increasingly realistic. The advantage of a predictive machine learning model is that it can be objectively assessed based on a blind hold-out data set, so you have clear criteria as to whether one model is ‘better than another.

What you need for an in-house approach

First, you need data and lots of it. Fortunately, digital marketing is one of the most data-rich activities on the planet. Chances are that if you are a B2C online business, you have hundreds and thousands, or else millions of site visit logs generated every week. These logs contain referrer data which helps you analyse and understand the customer journey to sale, the first essential step in building your attribution approach.

Second, you need a solid experienced data science and marketing analytics team. Go for a mixture of skills. Typically, some data scientists are strong on theory, but weaker on the application, while others are great communicators and see the strategic angle but are weaker at data-related programming. You also need domain expertise in marketing analytics. You need visualization experts and data engineering experts. The fabled ‘unicorn’ data scientist is impossible to hire, so instead, you should go for a team with the right mix of skills, with strong leadership to move the project forward.

Third, you need patience. The truth is, getting to an attribution model using machine learning is not easy. It is not a case of throwing some data at a model and waiting for the answers to pop out by magic. Your team needs to decide what data to use, how to configure it as a feature set, what types of model to use, what an appropriate training set is, how to validate the model and so much more besides. You will need to make multiple attempts to get to a half-decent model.

Choosing the final model

The best candidate ML models depend on your data – we have had good results with well-optimized classifiers and regression models, which we find often outperform even high order Markov models. While a black box or ensemble method may get better predictive accuracy, you need to consider the trade-off in terms of reduced transparency and interpretability. The best advice is not to commit to a particular modelling approach or feature set too early in the process, but to compare multiple methods.

But what then? An advanced machine learning model does not speak for itself. Once you have a model, you then need to be able to interpret it in such a way as to be able to solve the marketing mix allocation problem. What exactly is the contribution of each channel? What happens if spend on that marketing channel is increased or decreased? How does the model handle the combined effects of channels?

All of this will take months, so it is a small wonder that many companies ignore the problem or else go for a standard vendor out-of-the-box approach. It’s worth remembering then that there are some key benefits of a, ‘do it yourself approach to consider.

Benefits of an in-house model

If you create your model, you will discover a great deal about your marketing activity and data in the process. This can lead to immediate benefits – for example with one major international services firm we worked with we found significant marketing activity occurring in US states and cities where the company has no local service team. Even with no attribution model defined at that stage, the modelling effort uncovered this issue and saved the company huge sums right away. The point is that your data quality is tested and will become cleaner and more organised through the process of using it, and this, in turn, supports all your data-driven marketing.

Another beneficial side effect is that if you create your attribution model you will also learn about your business and decision making. This process will force your marketing teams to collaborate with data scientists and engineers to work out how to grow sales. Other teams need to be involved, such as finance, and your agencies, and this will often spawn further opportunities to learn and collaborate across all these marketing stakeholders.

Attribution is all about how different marketing channels work together, so your various marketing teams and agencies need to collaborate as well – social, search, display as well as above the line, and brand and performance more broadly. Again, this provides intrinsic and additional value over and above the modelling itself.

Finally, it is worth pointing out that you will never actually arrive at the final model. This is quite a fundamental point to bear in mind. By its nature, a machine learning approach means you need to train the model on fresh data as it comes in. Your marketing and your products are also changing all the time, and so are your customers. So really you need to build a marketing attribution modelling process more than you need to build a single attribution model.

So, go ahead, build your model, be less wrong than before, and then when you have finished, start all over again. As we say at Metageni, it is all about the journey.

Gabriel Hughes PhD


Can we help unlock the value of your analytics and marketing data? Metageni is a London UK based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com

Just how bad is your analytics data?

If you do not know how bad your analytics data is, then the chances are, it is much worse than you think. With data analytics, it is not the known data quality issues that will cause you the most trouble, not the known unknowns, but the ‘unknown unknowns’ – those issues you uncover and discover as you explore and analyse your data.

Usually, it is only the practitioners who are very close to the data who understand the full extent of the data quality problem. Too often the poor quality of data is kept as something of a dirty secret not to be fully shared with senior management and decision makers.

Common issues in web and marketing analytics

Let’s look at just some of the most common issues affecting web and marketing analytics data. To begin with, do not assume that the data sources provided by the most common analytics solutions are robust by default. Even the best ones are prone to big data quality issues and gaps. Take Google Analytics referrer traffic, which often reports an unrealistic level of ‘Direct’ traffic, supposedly visits made directly through users typing in URLs, or bookmarking, both low frequency methods of site access. The reason is that ‘Direct’ is, in fact, a default bucket used where no referrer data is available to the analytics server.

Experienced web analysts know that high levels of direct traffic usually mean high levels of broken or missing tags, or other technical issues, that means the true referrer data has been lost.

The major providers are also contributors to that other major source of poor data quality, which is fragmentation and disjoint data sources. Google search marketing tags will track conversions, but only from the Google search and display network. Facebook similarly provides tags which only link Facebook marketing to sales, ignoring all other channels. Affiliate networks do the same thing leading to widespread duplicated over attribution of sales to multiple sources. This challenge is exacerbated by different marketing attribution platform look back windows and rules which are different between platforms.

Having worked with multiple brands of all sizes, I have yet to come across a brand that does not have some level of tagging issue. A typical issue is a big mismatch between emails delivered and emails opened and clicked. Another is social campaigns which are delivered by 3rd party solutions and then appear as referral sources, due to the use of redirect technology.

Tagging and tracking

Tag management systems help manage this, but unfortunately not by linking the data, just by de-duplicating tag activity at source, which is hardly satisfactory if your goal is to understand Multi Touch Attribution (MTA) and marketing channel synergy.

Assuming you solve all your tagging issues and have well-structured soundly applied tags, you should not forget that the tag is only as good as the tracking itself. A great challenge here is the gap that exists tracking users across devices. You cannot link visits by the same user on different devices without sophisticated tracking that users have signed up to beforehand. This means your tags cannot tell the difference between the same user visiting twice on two different devices and two different users.

The idea every one of us can be closely tracked and monitored online is an illusion for all the biggest technology companies – and perhaps we should be glad of that. Indeed, unique ID tracking and linking is now more closely under scrutiny the age of data security breaches, increased concerns over user privacy and the GDPR. This is yet another source of difficulty for companies looking for a 360-degree view of the customer. Companies have to work with fully consented and well-defined samples of data to make progress in understanding their customers.

For the analyst, this is yet another reason why having huge volumes of data is not enough for user insight and data-driven decision making.

So what can you do about all these data quality challenges?

Data quality is perhaps like muscle memory in sport. You use it or you lose it. It’s only by trying to analyse and find patterns in your data that you uncover the issues that need to be addressed. Where there is a need, strategies can be devised to manage these gaps in data quality and take steps for improvement. It is a process.

The best advice is to get stuck in. Pick one data source and run with it, making sure to compare it to others and ask if the data makes sense given what you know about your customers. There are always discrepancies between data sources which in theory should report the same numbers: in my experience, this is a kind of law of all data analytics, so you need to get used to it. Use these differences to help you validate your sources, understand why differences might arise, and just accept that there is an acceptable level of difference – say 2-3%.

In data analytics, as in life, you must not let perfect be the enemy of the good. Be wary of the massive data technology project which promises to link all data together in one big data lake and thereby solve your challenges. Bad data plus more bad data does not equal good data. Face up to your terrible data quality, and tackle the ugliest issues head-on. If you ignore the problem, it can only get worse and you will continue to struggle forwards in the dark.

Gabriel Hughes PhD


Can we help unlock the value of your analytics and marketing data? Metageni is a London UK based marketing analytics and optimisation company offering support for developing in-house capabilities.

Please email us at hello@metageni.com