What spillover actually is
Spillover happens when a customer interacts with ads on multiple channels before converting, and more than one platform claims full credit for the sale. A customer sees a Meta video ad on Monday, searches your brand name on Google on Wednesday, and converts via a Google Shopping click on Friday. Meta claims the conversion because the purchase happened within its 7-day click window. Google claims it because the customer clicked a Google ad immediately before purchasing. You have one sale and two platforms each reporting 100 percent of the credit.
This is not a technical error you can fix by adjusting settings. It is the fundamental design of last-touch and view-through attribution models. The only way to measure true channel contribution is to step outside any individual platform's reporting.
How we measure spillover at Layer Insights
We use a four-tier probabilistic matching approach. The first tier is deterministic: customers who are identified by email or customer ID across both touchpoints are matched with certainty. The second tier uses device fingerprinting signals to connect anonymous sessions with high confidence. The third tier applies time-proximity matching: sessions within a defined window before a conversion on a different channel are flagged as likely spillover candidates. The fourth tier is a baseline correction using your channel mix to weight the results.
The output is a spillover coefficient per channel pair. A coefficient of 0.4 between Meta and Google means 40 percent of conversions claimed by both are likely the same purchase. We express this as a revenue range, never a single number, because probabilistic matching has inherent uncertainty.
The budget decision this changes
Once you know your spillover coefficients, two budget decisions become clearer. First, you can identify which channel pairs are cannibalizing each other's measurement most severely. A Meta to Google spillover coefficient above 0.5 suggests you are over-investing in one of them relative to its true incremental contribution.
Second, you can calculate your blended incremental ROAS across all channels rather than summing up platform-reported ROAS numbers that double-count the same purchases. A brand reporting a blended ROAS of 6.2 across three channels often has a true incremental ROAS closer to 3.8 when spillover is removed.
What to do with high spillover channels
High spillover between two channels does not automatically mean you should cut one. It means you need to run an incrementality test to determine which channel is driving the conversion and which is riding along after the decision was already made. Brand search on Google, for example, shows extremely high spillover with every other channel. This does not mean Google search is not contributing. It means most customers who convert via branded search would have converted anyway after being influenced elsewhere.
The right response to high spillover on branded search is to reduce branded search CPCs, not eliminate the channel. You are paying for customers who were already coming. That spend needs repricing, not elimination.
How to report this to your media agency
Most agencies report per-platform ROAS because that is what the platforms show them. Bringing spillover data into an agency conversation changes the framing from "which channel performed best" to "which channel drove the most incremental revenue." These are different questions with different answers.
Share the spillover coefficient between channels they manage. Ask them to separate brand search from non-brand search in their reporting, since spillover on brand terms is structurally different from spillover on category terms. Request that incrementality tests become a quarterly line item in the retainer, not a special project.
---