Paid Social Media Attribution in a world without Third Party Cookies

What’s the problem?

The big issue with understanding the value of paid social, display and video is that these channels tend to drive value in a passive and indirect way. People view these ads and develop awareness, opinions and expectations about the advertised brand, without necessarily doing anything that can be tracked online. In other words, they do not click on the ads, but the ads still have a psychological impact that can drive a sale later on.

This means that if you look at your Google Analytics at website traffic,  you will see very little in the way of visits directly from people who click on the ads before visiting the website. If they do not click, your analytics will not show that media as the traffic source. 

In other words you have a huge problem of ‘mis-attribution’ for these channels, whereby the visits you get from social media marketing actually appear as visits from other channels, such as Direct, or Brand Search.

It works like this: first, a potential buyer sees your ad, but does not click on it. Then, a bit later (could be minutes, hours or days) they are looking for a product that you sell, and they remember your ad or brand, and so they do a search and click on that to visit your site. None of the metrics listed above will attribute to your social media –  traffic, leads, conversions and revenue – they all show up attributing increased activity back to the search click.

The ‘view through’ problem is not new…

Even before the huge growth in social media and video, this was a problem for display ads, where click through rates have been falling for decades, even while ad spend on the channels has grown. Advertisers are used to spending money on activity like TV and print, and so are familiar with the idea that ads can generate brand impact which is hard to measure. Online we expect more, so as online becomes more like TV, some have said that measurement is going ‘back to the future’.

What has changed in recent years is that one major measurement trick has been removed from the toolbox. This is the 3rd party cookie tracker, which has fallen into disuse following concerns around privacy. Until Apple released ios14, and various browsers changed their cookie settings, it used to be possible to track people who saw your ad but did not click, and actually link into the site traffic and sales, even  if they did not click. This is no longer possible to the extent it once was and once Google completely remove third party cookies from their advertising system in 2024, these cookies will finally disappear.

The rise and rise of modelling

In the face of this technical shift in ad tracking, and with a growing need to measure Social and Video in particular, statistical and machine learning modelling has come to the rescue, combined with an increased awareness and availability of tools for incrementality testing (‘uplift tests’).

In the end, we have to accept that any kind ‘model’ of what works has to be better than flying blind. What is ironic is that this modelling done well can actually be more accurate than the previous method of cookie based attribution.  

With cookies and tracking for Social and Display the problem was almost in reverse – there was over-attribution to impressions! Channels like Facebook/ Meta had to use a 1 day ‘look back’ window to help explain the huge numbers of sales apparently driven by customers who have only briefly viewed 1-2 impressions before making a purchase in the past 24 hours. Any longer would have led to attribution that was just not believable. But did the ads really cause the sale or not? All the hard questions about real sales impact, brand vs performance, and marketing incrementality, are not solved by tracking cookies alone.

So, in some ways the end of cookie tracking will actually be a benefit to analysts trying to measure social media ROI, by taking away the noise of ‘post impression’ conversions, or so called ‘view-throughs’.

Today, for impressions generated by social and display, at Metageni we use econometric  approaches to analyse how the reach and frequency of impression campaigns interact with lower funnel marketing and sales, to model patterns in how the impressions influence sales over time. This can be baked into click based predictive attribution using machine learning, to produce a credible and accurate full funnel view of all digital marketing channels.

We recommend backing these up with uplift tests, and support geo based testing as a way to validate social media ROI.

This is how we help schuh

We help schuh, the leading footwear brand, get a better understanding of the impact of their paid social and display advertising through Meta, YouTube and TikTok. 

Metageni data scientists combine econometric techniques to complement the impression model with a custom machine learning attribution model which is developed for each client based on their marketing approach and unique data. 

Applying this approach for schuh has allowed them to gain a detailed understanding of how Social, Video and Display marketing works together with PPC, Affiliate and Organic channels to drive sales. This technique is used to make adjustments to performance click-led sales reflecting modelled impression effects and including ‘carry-over’ effects from one period to the next, allowing the measurement to pick up hidden responses to Social and Display campaigns which otherwise would get attributed to Direct, Organic and Brand Search. 

Metageni’s custom attribution platform has proven to be a valuable tool for schuh, allowing them to get the most out of their digital marketing campaigns. 

The results have been impressive, allowing a comparison between ‘click only’ and ‘impression driven’ attribution, with the combined effects attributing almost 25 times more sales than the clicks alone. Social Prospecting and Video in particular have much greater attributed sales and revenue, with a reduction in sales attributed to Brand Search, Direct and Organic Search traffic, of around 20%. With Video, the fact that so few users click on ads is a challenge for attribution, where it is clear that an increase in Video views driving a statistically measurable increase in site visitors who convert. Social Prospecting meanwhile gets the best of both worlds, with roughly 50% of visitors buying from clicks to the ads and another passive 50% coming to buy just after seeing ads but not clicking.   

The custom attribution platform provided by Metageni thus not only enabled schuh to optimise their digital spend for greater efficiency and ROI in PPC and other performance channels, but it also allowed them to measure the effectiveness of their social and display campaigns including the all important impression effects

Caroline McFarlane, Head of Digital Marketing at schuh, said – 

“Working with Metageni has been a game changer for our digital marketing. We’ve been able to better understand how our social and display marketing works together with our other channels to drive sales.

We know that our audience are avid social media and YouTube users, however using last click attribution we weren’t able to see the true value of these channels. By factoring in impressions we get a more rounded view of the role of each channel and its position in the customer’s path to purchase, allowing us to optimise our digital spend for greater efficiency and ROI.”

Brands are now left with little choice in 2024 to get an accurate reach on their digital spend

If it sounds tricky, well frankly it can be, but brands are now left with little choice in 2024. If you are reading this and looking for an easier way into it, then just remember to separate out the click effects form the impression effects and ask yourself all the same questions you might ask if you were advertising offline – 

  • Who is my audience?
  • What do I really expect them to do in reaction to this campaign? (clue: not necessarily purchase immediately they see the ad)
  • What proportion of the audience will response over days and weeks vs immediately (clue: most of them!)
  • How do I account for the hidden impression and reach effects which underpin most of the value? (clue: you cannot track it, you cannot ignore it and so you need to model it!)
  • How many times will they see my ads and how does frequency of exposure help drive greater value? 
  • How much of the lower funnel sales activity is actually driven brand awareness generated earlier?

In other words, think about your audience and customers, and how they actually behave, when you measure your impact for social media. You will soon come to understand that true ROI is much higher than just a few clicks on your social ads.

If you are struggling to measure the ROI of your social media marketing efforts, and wish to learn more about how custom attribution and marketing effectiveness could help transform your campaign success, then get in touch.


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 and custom attribution using machine learning, or are interested in how Metageni can help you use your data to grow online, then do get in touch with us:

And check out our free guidebook on marketing measurement.