Facebook Match Rates

To determine the match rate range for a Facebook custom audience created via LiveRamp, first divide the Facebook custom audience low range estimate (given within the Ads Manager UI) by the number of input records in the field or segment you distributed to Facebook (on the Audience Segments page in LiveRamp Connect). Then do the same with the Facebook high range estimate for that audience. This will give you a match rate range, from the lower estimated number to the higher estimated number.

When calculating match rates for Facebook, LiveRamp does not use the number of records (rows) sent to Facebook in the calculation; instead, the number of input records from the original field or segment (also known as “LiveRamp Input Records”) is used.

LiveRamp does not use the number of records (rows) sent to Facebook because oftentimes multiple rows of information are sent to Facebook (if using the Direct Integration) or multiple ExternIDs (if using the Managed or Data Marketplace integrations) per original input record. Using the number of rows (rather than LiveRamp input records) would cause Facebook’s match rate to be misleading and wouldn’t accurately reflect the original 'size' of the field or segment.

You can view the input records stat on the Audience Segments page in LiveRamp Connect, but unfortunately this information is not directly visible in Facebook or within the Ads Manager UI.

It's not abnormal to see lower than expected match rates when delivering data to Facebook, due to the way the LiveRamp integration with Facebook works. We will often distribute more than one row of data per “person” in the original file, and then Facebook will match those multiple rows back to one Facebook user, as expected, but because of how the logic for Facebook’s “low match rate” flag is designed (it will appear if Facebook matches less than 30% of the API calls sent to an audience), the Facebook UI will show a “low match rate” flag even though the match rate is not actually low.

For example, we may distribute 5 rows of identifiers for a single “person” from the original file, and then Facebook will match these 5 rows back to a single Facebook user, but their matching logic erroneously calculates a 20% match rate (1 user/5 rows), even though the actual match rate was 100% (1 user/1 input record).

If you calculate the actual match rate by taking the audience size from Facebook and dividing that by input records from within LiveRamp, you will typically find a ‘normal’ match rate.”

There are several factors that can contribute to a higher than expected match rate at Facebook, in addition to some of the typical reasons match rates are sometimes higher at destination platforms (such as the precision level used).

Due to the nature of our Facebook integration, LiveRamp does not know which set of PII a person has shared with Facebook. Therefore, LiveRamp may send the same person to Facebook multiple times with different combinations of PII which may match to different Facebook users. This optimizes for reach, ensuring we match the target user on Facebook.

Due to the nature of our integration with Facebook, LiveRamp is unable to receive a match rate from Facebook and therefore, Facebook match rates fall outside of how LiveRamp typically defines match rates. The Facebook match rate is Facebook ID (equates to Facebook account, not user) over LiveRamp input records. This can sometimes cause anomalies in match rates when a user views their match rate within Facebook as Facebook is unable to receive LiveRamp input record counts.

CID Data does not expire at Facebook: When an audience is first uploaded to LiveRamp, we send it over to Facebook. If you subsequently fully refresh the audience, we send over any new users you have added to the audience, However, older users no longer in the audience do not expire and will remain in the audience count on the Facebook side. This can cause an increase in match rates.

Including multiple identifiers for each person: Having multiple PII in a record can increase match rates significantly as it increases the number of RampIDs we match for each record, effectively increasing the audience size of CIDs we send to Facebook.

Multiple Facebook IDs: Some users have multiple Facebook accounts for either fun or work. In cases such as this, it becomes likely that Facebook will match one of the user accounts when we send some hashed-PII and the other user account when we send a different combination of hashed-PII. When the hashed PII we send matches to multiple Facebook users, it will lead to a match rate that is over 100%.

The “low match rate” flag will appear if Facebook matches less than 30% of the rows sent to an audience in Facebook. This often occurs due to how we distribute data to Facebook.

Before we distribute the data to Facebook, we break out each individual identifier in a record into a separate row before matching each row, first to a and then to the corresponding Facebook CIDs. For example, for one input row that contains name and postal, phone 1, phone 2, email 1, and email 2, we would create five rows, one for each individual identifier. Because of this, we will often distribute more than one row of data per “person” in the original file.

Facebook matches CIDs to Facebook users using a mapping they hold of CIDs<>Facebook users. This mapping is refreshed once a quarter, when we send Facebook a new mapping of CID<>hashed PII, which they use to create the mapping of CID<>Facebook user on their end.

If Facebook then matches those multiple rows we sent back to one Facebook user, because of how the logic for that “low match rate” flag is designed, the Facebook UI will show an artificial “low match rate” flag even though the match rate is not actually low.

In this example, if Facebook matches those five rows back to one Facebook user, Facebook calculates this as a 20% match rate, even though the true match rate was 100% (one input record to one Facebook user).

If you calculate the actual match rate by taking the audience size from Facebook and dividing that by the number of input records from within LiveRamp, you will typically find a ‘normal’ match rate.