Making Ad Metrics Clearer


Facebook heard that businesses want more insight into our measurement tools and metrics. So today, we’re introducing new labels on some of our metrics to clearly show how they are calculated, and we’re removing some other metrics to help you focus on the more meaningful ones. We're also launching a program to help educate marketers more broadly about measurement.

More transparent metrics

Starting today, we’ll begin labeling some metrics in Ads Manager as estimated or in development, to provide more clarity on how they are calculated and how you should consider using them. These labels will appear in tool tips within the Ads Manager reporting table and in the customize column selector for ads running across Facebook, Instagram and Audience Network.

Estimated metrics are calculated based on sampling or modeling. They can provide guidance for outcomes that are hard to precisely quantify, and we may update our measurement methodologies as we gather more data and improve our signals. When we provide real-time results, we often use sampling methods that allow us to instantly model metrics at scale. By labeling metrics as estimated, you will now know when these methods are used.

For example, reach is an estimate of the number of people who saw an ad at least once. In order for us to report reach, we analyze the number of people who see an ad multiple times, de-duplicate them and then calculate the total number of unique people in real time. To do this quickly, we sample the data and will therefore label it as estimated. This is also how reach is calculated for ads on TV and across other digital platforms.

Metrics in development may be new or in testing, so they may evolve as we improve our ad products and measurement methodologies. When we launch new ad features, they require metrics that we must test to determine how to provide the best insights. Because the metrics may change as products evolve or as we get more feedback from businesses, we’re indicating these are still in development and subject to change.

For example, estimated ad recall lift is a metric used by brands to understand the differences between people who can recall a brand after seeing an ad compared to those who have not seen an ad. This kind of automated measurement is still new and requires both polling and machine learning. Because we use sampling to determine this metric, it will be labeled as estimated, and since we’re still gathering advertiser feedback on it, it will also be labeled as in development.

Removing some unhelpful metrics

In July, we will remove approximately 20 ad metrics that marketers have told us are redundant, outdated, not actionable or infrequently used. For example, the social reach metric shows the number of people who saw an ad with social information above it, such as noting a friend who also likes a certain brand. We’ve heard from you that this metric isn’t meaningfully different from the reach metric, and we know that the insight drawn from it doesn’t indicate a business outcome.

Removing these types of metrics will make it easier for you to get the most actionable insights to improve your ad performance. Visit the Advertiser Help Center to see the full list of metrics that will be removed.

Measurement education

Even as we improve and clarify our metrics, we know businesses still need help understanding broader measurement principles. That’s why we’re introducing a new program called Measure What Matters that will be rolling out in March and will offer two tracks—one for advertisers focused on brand objectives and one for advertisers with direct response objectives. Each track will draw from research and analysis across creative planning, ad delivery, cross-channel measurement and video measurement, and we'll share these insights with you through in-person events, Facebook Live events and on the Facebook Business website.