Estimated Ad Recall Lift Rate: Can it be Trusted?

Estimated ad-recall lift rate is a lesser used metric for me in Facebook Ads Manager. However, I’ve started to use it more when trying to understand the performance of brand awareness campaigns and the performance of audiences and assets within those campaigns - but it comes with a caveat.

What is Estimated Ad Recall Lift Rate?

Estimated ad-recall lift rate as a calculation is: Estimated ad-recall lift (people) divided by Reach, so it is represented as a percentage. It’s basically an estimated percentage of people who would recall seeing your ad if they were asked within two days - a bit of guidance as to how memorable your ads are.

Can I Trust Estimated Ad Recall Lift Rate?

• The metric is estimated, as obviously stated in the name. This means that it will never be a precise figure. Meta uses sampling and modelling to calculated estimated metrics:

  • Sampling means they take a smaller portion of the data, and then apply the results of that data across the whole dataset.

  • Modelling means signals are used to help with the calculation, such as ad interactions, or even takes data from similar campaigns.

  Estimated metrics therefore should be used as guidance only, and not be reported as fact.

• There is actually a huge list of metrics which are labelled as being ‘estimated’, including Reach, Unique Clicks, Purchases, to name a few (and thus all the metrics that use those in their calculations). I think most people typically report on Reach as being factual, so…

• The metric is also in development, which makes it sound like it’s in Beta, and has been for years. Metrics in-development mean they are still being tested on, and Meta is still working out the best way to report on them. From research, metrics using machine learning will technically always be in-development due to the very nature of machine learning. The list of metrics in-development is also hefty, but include much rarer metrics.

• There are some metrics, particularly engagement metrics, that are subjectively good or bad - we don’t necessarily understand the performance of them until we compare them to other activity. The same can be true for Estimated Ad Recall Lift Rate; if it’s calculated in the same way the majority of the time, then at least we can understand and report on the metric relative to other ads or targeting alongside each other.