Learn how an automotive brand tested a proposed Google Brand Search strategy against its existing one, at the same spend, before moving any budget.
A 42-day, three-cell geo incrementality test showed that an automotive brand’s existing Google Brand Search strategy returned $31.03 in incremental revenue per dollar, nearly five times its model’s estimate. A proposed new strategy, tested at the same spend, returned $14.27. Switching would have cost an estimated $31.7M in incremental revenue a year, so the brand kept its existing approach. Here’s how the test worked.
Business Challenge
Google Brand Search was the brand’s third-largest channel by spend, and the team was weighing a new strategy for one of its vehicle model lines. Brand Search is hard to value. People searching for a brand by name were often going to buy anyway, so a reported return on its own says little about how much revenue the channel really adds. The brand’s model put the return at $6.55 per dollar. Changing strategy on a channel this size without evidence was a risk leadership wasn’t willing to take, so they asked for proof of which approach drove more incremental revenue.
Objectives
Measure the incremental revenue from the existing, business-as-usual (BAU) Brand Search strategy.
Test the proposed new strategy head to head against BAU, at the same spend, before shifting any budget.
Use the results to correct the model’s estimate for Brand Search.
How LiftLab Tested Two Brand Search Strategies Side by Side
The team designed a three-cell geo experiment, so both strategies could be measured against the same baseline over the same period.
Three matched cells: For 42 days, markets were split into three matched cells of 22% each: a control cell where Brand Search went dark, a BAU treatment cell and a new-strategy treatment cell.
Equal spend: Both treatment cells averaged $5,100 a day, so any difference in results came from the strategy, not the budget.
Go Dark with Pacing design: The control cell gave a clean baseline for what sales looked like with no Brand Search at all.
Calibration through the Trust Engine: The results were fed into the brand’s Agile MMM, correcting its Brand Search estimate.
Results & Business Impact
The BAU strategy returned $31.03 in incremental revenue per dollar, 373% above the model’s pre-test estimate of $6.55. The existing strategy was working far better than the model gave it credit for.
At the same spend, the proposed new strategy returned $14.27, less than half of BAU’s return.
BAU drove a 4% lift (97 incremental orders), compared with a 1.5% lift (37 orders) for the new strategy, over the same window.
Annualized, the gap between the two strategies was large.
BAU’s incremental revenue annualizes to about $58.6M, against $26.9M for the new strategy.
Switching fully to the new strategy would have given up an estimated $31.7M in incremental revenue a year.
The brand avoided a costly change and now has an experiment-calibrated benchmark for one of its largest channels.
[Fascinating stuff, right? Here’s another story of how a customer uncovered nearly $14,000 wasted daily Google ad spend with LiftLab.]
Key Takeaways
Test strategy changes before you roll them out. A three-cell design measures the current approach and the proposed one against the same baseline, at the same spend, so leadership can choose based on evidence.
Brand Search incrementality is brand-specific. Brand Search is often assumed to capture demand that would have arrived anyway. For this brand, it drove substantial incremental revenue. The only way to know which is true for yours is to test it.
A model’s estimate is a starting point. The pre-test model valued Brand Search at less than a quarter of its measured return. Calibrating with experiments keeps the model honest in both directions, as a related Meta Awareness test also showed.
Read small counts with care. The difference rests on 97 versus 37 incremental orders over six weeks. Look at the confidence range alongside the headline figure, and re-test before applying the result beyond the model line tested.
Methodology note: iROAS (incremental return on ad spend) is the incremental return observed during the test window, comparing each treatment cell with the control cell, where Brand Search went dark. Annualized figures extend the test-window results to a full year at the tested spend level. Results reflect one anonymized customer experiment run in Q3–Q4 2026. Individual results vary by category, baseline demand and media mix.
Our customers’ names are kept anonymous to protect their confidentiality.
Is your Brand Search strategy earning what you think it is?
FAQs about the Google Brand Search Case Study
Is Google Brand Search incremental?
It depends on the brand. Some of the people who click a Brand Search ad would have bought anyway, through an organic result or a direct visit. A geo test that turns Brand Search off in some markets shows how much revenue it actually adds for your brand.
What is a three-cell geo test?
It splits markets into three matched groups: one where the channel goes dark and two that each run a different strategy. Comparing each strategy’s group with the dark group shows the incremental return of each, measured over the same period.
What is the Go Dark with Pacing methodology?
Go Dark with Pacing is LiftLab’s experimental method for measuring incremental media impact. It pauses delivery in selected control markets while maintaining normal delivery elsewhere, allowing measurement of the difference in business outcomes.
Why would a marketing mix model underestimate Brand Search?
Brand Search volume tends to rise and fall with underlying demand for the brand. That can make it hard for a model working from historical data alone to separate the channel’s effect from demand that was already there. A live experiment measures the effect directly.
How does LiftLab use experiments to calibrate its MMM?
LiftLab’s Incrementality Testing Suite runs geo experiments like this one, and the Trust Engine feeds each result into Agile MMM to update the affected channel’s estimate and response curve. Every budget recommendation that comes out of the model goes to your team to act on.





