TLDR: If you’re measuring performance of ads based on the conversion data Facebook is showing you, you’ve been optimizing using unreliable data & not making the best decisions to reach your goals. Not only are you missing out on attribution, but you’re also getting misattribution. We have a fix to help you make the right adjustments.

So you’re probably wondering, misattribution?

Attribution is a hotly discussed topic lately, which even led to some pretty aggressive Twitter chats in the last couple of weeks. Because of the limited data Apple is passing back to Meta, conversions are being attributed to the wrong ad, the wrong ad set, or the wrong campaign. For this post, I’m going to stick to the ad level, although there are other areas affected by this issue. The ad context is the most severe and is where implementing the current solution we’re using at Elumynt can have the biggest impact.

Meet: Aggregated Data Dumping

Across clients, we’ve been seeing that conversions have been attributed in bulk. This is to anonymize the conversion data so it cannot be traced back to an individual, but the problem this poses for advertisers is not truly knowing which ad is driving your conversions.We originally thought this wouldn’t significantly affect our decision making, as we assumed the conversions being traced back to ads would remain at a relatively similar percentage to how Facebook places them in platform. But you know what they say about assuming…After comparing Google Analytics’s last click reporting, Google Analytics Path's Report, and in-platform Facebook attribution (using 1-day click), we found the proportion of conversions from different ads within an ad set was drastically different between the two reporting platforms. There were instances where Google Analytics would show Ad A with 11 conversions and Ad B with 1 conversion, while Facebook showed Ad A with 3 conversions and Ad B with 24. That can’t just be explained by different attribution windows.See the conflicting results from this actual ad:

If you were looking at Facebook attribution, you’d say with absolute certainty the Red ad was the best performer, assuming spend and order value were constant. With that information, you’d turn off the blue ad, confident you were making a data-backed optimization. But when you look at a different source (Google Analytics), you’d find the Blue ad is actually driving more conversions. In this instance, Google Analytics is arguably more trustworthy since it's looking at the actual user and capturing the UTMs from the specific click/visit and then assigning that the conversion. It could end up with less totally reported conversions (which is what we see in the example above, with 3 vs. 6), but it's at least accurate with what it did track and not as impacted by the aggregated data dumping issue.Every day, advertisers are making the wrong decisions because they don’t understand what iOS14 did to the data they’re looking at.

In practice, this means Apple passes back a "load" of conversions in bulk to Facebook, which then Facebook "dumps" on a single ad as it sees fit. Since the data is anonymized they can't parse it out to the actual converting ad with any level of certainty so they just slap it on one they deem is a high performer.

So…what do we do now?

To look at the data like we do, you need to have campaign, ad set, and ad name included in your UTM parameters in the ads.utm_source=facebook&utm_medium=paidsocial&utm_campaign={{campaign.name}}__{{ca

mpaign.id}}&utm_content={{adset.name}}__{{adset.id}}__{{ad.name}}__{{ad.id}}Then you can create a report to see last click attribution that enables you to have a more accurate and holistic view of Facebook's actual performance. Here is the custom report we’ve built in Google Analytics :

This report will tell you which ads are driving last click attribution, which should at the very least be indicative of which ads are top performing and tell you which ads aren’t performing (at least via last click). Again, this method won't give you ALL of the conversions, but it will help you better understand the trend between which ads are doing a better job of driving actual sales based on clicks.*Note: in this example report, we are looking at ANYTHING that's attributed to "facebook" at the source/medium level, not just "facebook / cpc" or something like that. We've found this to be more reliable for looking at the bigger picture for many brands, but if you have a lot of organic purchases from Facebook (i.e. VIP Facebook Group, etc.) it might be worth it to dial this in more specifically. This also misses the attribution from Instagram, which you may or may not need to include depending on how much volume your business does on Instagram.

Results May Vary

That said, the severity of this depends on the business. If a company is bringing in hundreds of orders per day, the conversion “dumps” are larger in volume and more frequent, being attributed to any number of ads that are live.For a business that has fewer orders per day, they may have smaller dumps of just a few conversions at a time or may still come as a single conversion attributed to the correct ad. These companies also have less frequent dumps, which makes them more prone to attribution coming through a couple of days later.

This, in addition to Elumynt’s Growth Strategy Playbook, has allowed our clients to continue exponentially scaling growth despite our industry’s struggles of the last year.

Remember that Facebook’s attribution is changing every day, and our processes for evaluating and optimizing ads will continue to evolve at the same rate. This is one solution we're using along with several other attribution models now, but we may have a better solution tomorrow for ya 😉. There are a lot of attribution tools that claim to use A.I. to correctly attribute things across accounts, campaigns, ads, etc., and I won't weigh in on those at this time, but just know that those can often show significantly different data between each other. Those solutions, however exciting, are unproven compared with tracking attribution directly from clicks in Google Analytics. Additionally, it's good to use post-purchase surveys and other data to come to a full picture of attribution which is something we discuss at length when we chat about ROAS vs. MER.