Facebook attribution: chicken or egg?

Does your remarketing really drive incremental revenue?

We regularly carry out media audits for clients which involves analysing conversions generated from various digital marketing channels. Attribution in other words.

 

When performing this analysis, we typically look at web analytics reports like this one from Google Analytics.

Google Analytics default channel grouping report

Judging by the numbers, Paid Social, which stands for Facebook ads in this example, seems to be underperforming here. True, this is based on last non-direct click attribution. But also when looking at this from a multi-channel perspective and taking assisted conversion value into account the absolute revenue remains low.

Google Analytics multi-channel funnels report
Google Analytics multi-channel funnels report

This typically leads to discussions where the argument is brought forward that web analytics attribution is flawed because it doesn’t handle cross-device measurement as well. Which is particularly unfavorable to Facebook ads given the vast mobile user base Facebook has. Is this a valid argument?

 

Compared to other channels Facebook has a disproportionally high percentage of mobile users. In the example used this is as high as 90%. If your analytics tool is indeed notoriously bad at measuring cross-device traffic you do have an issue. If these Facebook ads are responsible for driving revenue in a significant way it would basically mean that users who are exposed to Facebook advertising campaigns would resurface on another device via another channel to complete the conversion and no conversion value would be attributed to Facebook advertising campaigns.

 

However, the mobile usage of other channels in this example tends to be in the range of 45%-60%. Not as high as Facebook, but still considerable. If these channels were equally affected by flawed cross-device measurement they would also be affected in a major way. Yet, these channels do show significant absolute revenue amounts, up to six figures even. And they are measured with the exact same tool in the exact same way with the exact same supposed cross-device measurement issue. Could it be the case that the cross-device issue is overstated and the reported Facebook advertising campaign revenue is low because, well, there simply isn’t any more revenue to report?

 

To offset this lack of conversion impact the industry has responded by pushing Facebook Attribution forward as an alternative. Facebook Attribution does not only measure Facebook, but also all other digital marketing channels. Different attribution models are available, some of which credit all channels in the conversion path, as opposed to just a single one. And to top that off, because of the enormous user base Facebook has, identifying users across different devices is a piece of cake using login-based identifiers. Let’s get this party started!

 

Not so fast. There is a catch with Facebook Attribution. The lookback window for attributing credit to ad clicks is 28 days by default (or shorter if you want), which is perfectly ok for most conversion scenarios. But, in addition, Facebook also applies a lookback window of 1-day to attribute conversions to ads that weren’t clicked but viewed. Let me repeat that: Facebook attributes success to ads that were viewed but not clicked in the last 24 hours. And that’s a big issue.

 

Even though a 1-day lookback window seems very short, just consider the immense user base Facebook has and the massive amount of ads served on a daily basis. It’s impossible to use Facebook without scrolling past ads that you do not click on. I probably open my Facebook mobile app 5-10 times per day and scroll past many ads that I will never ever click on. My timeline has evolved to a list of remarketing ads from just about every site I visit these days, whether for personal or professional purposes. Facebook Attribution will abuse this to make you as an advertiser believe that Facebook Ads have a massive impact on your conversions. If you care to dig a little deeper you will find this is well besides the truth.

 

Let’s illustrate this with a real-life example. One client ran a campaign for a specific product line on Facebook. Facebook Attribution reported 127K of revenue being generated by a Facebook advertising campaign alone. Whereas the client’s ERP system reported only 17K worth of this product line being sold in that same period through any channel. The graph below based on these ERP data confirms no more of these products were sold during the Facebook campaign periods than outside Facebook campaign periods. If Facebook Attribution claims a revenue of 127K then surely one would expect to see outspoken revenue peaks during campaign periods? Something smelling fishy here?

Lack of incremental sales from Facebook campaigns
Revenue per day from product line concerned (ERP based)

So why does Facebook claim 7,5x times more sales than there actually were?

The answer to this is twofold. First of all 70% of the revenue that was attributed to Facebook was based on ad views as opposed to clicks. This is criminal! Just think about all the ads you scroll past on Facebook on a daily basis. For all of those ads Facebook will be claiming conversion credit. What’s more, when an ad has been both viewed and clicked the revenue attribution is double counted, once for the view and once for the click.


Secondly, from historical data we were able to establish that 75% of sales of the product line involved in this campaign typically occur on the day of the first visit to the website. These are relatively simple and low-priced consumer products that are bought swiftly. However, Facebook attribution only reports 1,5% of product sales to these products being bought within the first day after clicking the ad.

Impact of Facebook view-through conversion in Facebook Attribution

The reason for this is that the advertiser concerned also sells many other and more complex and high-priced products, also to different customer segments. When users visit the website to view these products and subsequently buy them after one or more weeks that revenue is also attributed to the same Facebook campaign, which didn’t even feature these products! There’s a bit of pixel magic for you.

In summary

Does all of this culminate to the fact that switching off Facebook campaigns will have hardly any impact on your revenue? All cases are different but this might indeed well be the case. A large advertising client facing the exact same attribution dilemma actually ran a Facebook campaign with two different ads. One for the actual product (ad variant A) that was being advertised and one for a product that had nothing to do with the company (ad variant B). Both ads led to the same landing page. Sure enough, the attribution reports that came back also attributed a significant amount of revenue to ad variant B which had nothing to do with the company. Based on this finding the advertiser actually switched off all ads for months on end and their website traffic and revenue remained completely stable.

Moral of the story

Facebook Attribution reports are often presented as a silver bullet to back up increased investment in Facebook advertising campaigns. Simply apply the before-mentioned learnings: discount post-view revenue and analyse to what extent the remaining post-click revenue correlates with your own back-end product sales data. If you don’t do this chances are you will end up spending massive amounts of media budget on campaigns that barely have a direct revenue impact. But you will receive beautiful reports claiming they do.

 

Does this mean you shouldn’t be advertising on Facebook altogether? No. But approach it as a branding tactic rather than a performance marketing tactic. Let’s stop pretending these campaigns have a massive revenue impact unless there are objective ERP data to show for it.