Study: Last-Touch Ad Attribution Undervalues Facebook by as Much as 30%

If you’re using a last-touch attribution (LTA) model to measure the value of Facebook advertising, you’re undervaluing that channel by as much as 30 percent. This is the major finding of a new study from Kenshoo, “Quantifying the Impact of Multi-Touch Attribution: Spotlight on Facebook,” which analyzed clients’ live campaign performance data across multiple digital marketing platforms in March, April and May.

Background
According to Kenshoo, the brands included in the study came from a variety of industries, including retail, home improvement and financial services. The company notes that while multi-touch attribution (MTA) models can recognize the impact of ad impressions on conversions, for the purposes of the study only clicks were analyzed.

The study used cost per acquisition (CPA) across all advertisers included in the analysis. CPA is defined as the total media cost divided by the total number of conversions (e.g., filling out a form, an online sale, an app download, etc.).

Here’s how Kenshoo explains the crux of the study:

To understand how individual channel performance would be valued via Last Ad versus other standard attribution models, the CPA was first run through a Last Ad model and then those same customer paths were analyzed under various MTA models. For example, if the Last Ad CPA for a campaign was $10, and the multi-touch attribution model being compared reported the CPA as $9, it means the channel was being undervalued by 10% versus Last Ad. In other words, that channel was getting 10% less credit than it should have received for its role in driving conversions.

First Only
The “First Only” attribution model gives credit for a conversion only to the first click or interaction a customer has with the advertiser, regardless of whether the customer interacts with multiple keywords or channels. “First Only attribution values media interactions at the top of the funnel to make sure the most people are aware that the product/service exists,” according to Kenshoo.

The study finds that the LTA model undervalues Facebook by 30 percent versus this approach.

Prefer First
This attribution model gives most of the credit to the first ad, with subsequent touches receiving credit in a “rules-based, linearly decaying fashion.” This gives more credit to the touches that happen as the customer is pushed through the funnel.

Kenshoo finds that the Last Ad model undervalues Facebook by 20 percent versus this approach.

Divide Equally
As its name suggests, the “Divide Equally” attribution model gives all interactions equal credit for a conversion. This is a safe bet for marketers with intricate media engagements and long sales cycles, according to Kenshoo. “With each touch earning a bit of credit for assisting conversions, it becomes very easy to see which placements are not driving any value at all,” the report notes.

The LTA model undervalues Facebook by 16 percent versus this attribution model.

Prefer Last
This attribution model is the inverse of the “Prefer First” model, meaning that while the last touch gets the most credit, the previous touches get credit in a “rules-based, linearly decaying fashion.” This is a way for marketers to test multi-touch attribution models without moving too much away from Last Ad attribution, according to Kenshoo.

Last Ad attribution undervalues Facebook by 12 percent versus this approach.

U-Shaped
This attribution model uses an 80-20 rule