How to Combine Incrementality Testing and Multi-Touch Attribution for BFCM 2026 Budget Planning

On May 5, 2026, Google previewed Meridian GeoX, an open-source tool built to run geographic incrementality experiments and feed the results back into Meridian as calibration priors.

It’s a small technical announcement with a large implication: even the platform building the industry’s leading marketing mix model no longer treats attribution and incrementality as competing methods.

It treats them as inputs to the same system. That’s the exact combination BFCM 2026 budget planning needs, and most teams still run these two tools in separate silos instead of connecting them.

Why Neither Method Alone Is Enough for BFCM Budget Calls

Multi-touch attribution is fast. It updates daily, it credits every touchpoint a customer passed through, and during Cyber Five, when spend and traffic move by the hour, that speed is genuinely valuable for in-flight reallocation.

But MTA is still a model, not a causal test. It tells you which touchpoints a converting customer encountered. It can’t tell you whether that customer would have bought anyway.

Incrementality testing answers exactly that causal question, but it’s slow by comparison. A holdout test needs two to six weeks to reach statistical significance, which makes it useless for deciding whether to shift budget from TikTok to Meta at 11am on Black Friday.

Run it alone and you have ground truth that arrives too late to steer the moment that matters most. Run MTA alone and you’re making real-time decisions off a model that’s known to overstate certain channels badly.

Branded search and Performance Max, the two categories that spike hardest during BFCM, routinely overstate their platform-reported ROAS by 60 to 80 percent against measured incremental lift. Neither tool alone gets you through Cyber Five with numbers you can actually trust.

The Combined Framework: Where Each Method Does the Job the Other Can’t

Use incrementality before BFCM to set your baseline iROAS by channel

Run holdout tests on your highest-spend channels in the calm weeks before Cyber Five, while volume is still normal enough to reach a clean, reliable read.

The output isn’t just a report, it’s a set of guardrails: a baseline incremental ROAS for each major channel that you carry into BFCM week as the number you actually trust, separate from whatever the platform dashboard says.

Use multi-touch attribution during BFCM for daily, in-flight reallocation

Once those guardrails exist, MTA’s speed becomes an asset instead of a liability. You’re no longer asking MTA to tell you the causal truth.

You’re asking it to tell you where spend is moving today, bounded by an iROAS ceiling incrementality already proved is real.

If branded search shows a platform ROAS spike on Black Friday, but your pre-BFCM holdout already established that branded search typically overstates by 60 to 80 percent, that context changes the reallocation decision from “scale it hard” to “scale it modestly, and don’t confuse the spike with new demand.”

Feed BFCM results back into calibration, not just a report

This is the step most teams skip, and it’s the one Google just built infrastructure for. Meridian GeoX is explicitly designed to convert geo-experiment results into Bayesian priors that calibrate an MMM, so a single incrementality test doesn’t just answer one question and get filed away.

It permanently improves the model’s accuracy for every future budget call. AdBeacon’s own Meridian MMM integration follows the same logic: your BFCM holdout results and your BFCM MTA data both belong in the same calibrated system going into Q1, not two separate spreadsheets nobody reconciles.

A Practical BFCM Budget Decision, Walked Through

Say your branded search campaign shows a platform-reported ROAS of 9x on Black Friday, and your MTA model credits it as a major driver of the day’s revenue.

Taken alone, that’s a strong case to shift more budget into branded search for Cyber Monday.

But your pre-BFCM incrementality test already showed branded search running 20 to 60 percent above measured incremental lift on this account specifically, consistent with the wider pattern of branded search and PMax being the two categories that overstate hardest.

The calibrated view says something different: much of that 9x is demand that existed already, captured rather than created. The better reallocation is toward the channel your holdout proved actually creates demand, even if its platform ROAS looks less impressive on the dashboard that day.

That’s the decision incrementality and attribution make together that neither makes alone.Building this into your BFCM plan is what separates teams reacting to whatever dashboard looks best in the moment from teams making calibrated calls.

If you want help setting up first-party incrementality baselines ahead of this BFCM, feeding into a measurement system that actually connects the two methods instead of running them in parallel, book a live AdBeacon demo before spend ramps.

—-

FAQ

Should I trust MTA or incrementality more during BFCM?

Trust incrementality for the causal question of what’s actually driving demand, and use MTA for the speed of in-flight, day-to-day reallocation once incrementality has set the guardrails. Neither should override the other; the value comes from using each for the decision it’s actually suited to.

Can I run a new incrementality test during BFCM week itself?

It’s not ideal. Holdout tests need two to six weeks and a stable baseline to produce a reliable read, and BFCM’s traffic spikes and behavior shifts make that baseline hard to establish mid-event. Run your incrementality tests in the calm weeks before Black Friday, then use those results as the calibration layer once BFCM week arrives.

How does Google’s Meridian GeoX change this for Google Ads specifically?

Meridian GeoX is Google’s new open-source tool for running geographic incrementality experiments and feeding the results directly back into a Meridian MMM as calibration priors. Once it’s broadly available, it formalizes exactly the combined approach this article describes, using tested, causal results to correct what attribution alone would overstate.

What if I don’t have time to run incrementality tests before BFCM this year?

Use general benchmarks as a stopgap. Branded search and Performance Max typically overstate platform ROAS by 60 to 80 percent against real incremental lift; retargeting runs high too. Apply a discount to MTA’s credit for those specific channels rather than treating the platform number at face value, and prioritize running your first real holdout test in the calmer weeks after BFCM so you have it in hand well before next year.