The Branded Search Budget Everyone Wants to Cut
An e-commerce retailer spends $250,000 per month on Google Ads: 60% Performance Max, 15% branded search, and 25% non-brand search. Performance Max runs at a 400% target return on ad spend and is actually delivering 520%. Branded search holds 92% impression share, while non-brand impression share is 28% with 40% of impressions lost to budget. The finance team noticed that Performance Max is capturing many queries containing the brand name and has proposed eliminating the branded search campaign and shifting that budget into Performance Max, arguing it will lower cost per acquisition because Performance Max already converts so well. The paid media lead is uneasy but cannot articulate why, and has asked for an analysis before the next budget meeting.
1. Diagnose what is happening across these three campaign types and why the finance proposal is risky.
Read the worked answer
The proposal confuses correlation with causation in two places.
First, branded search is largely defensive. At 92% impression share, the campaign is protecting a query where the searcher already knows the brand. Cutting it does not create new demand, and competitors or retailers often bid on your brand terms, so some of those clicks would be captured by others. The correct test is a geo pause experiment, not an assumption that brand spend is waste.
Second, Performance Max looks efficient partly because it is absorbing brand and lower-funnel queries. When a campaign can serve on branded searches, its reported return is inflated by demand that other channels generated. This is the real issue behind finance's observation, and the fix is exclusion and transparency, not consolidation.
Third, the non-brand data contradicts the premise. Non-brand impression share is 28% with 40% lost to budget, meaning there is proven, unserved demand in the actual growth channel. Cutting a 92%-saturated defensive campaign to fund an opaque one while leaving non-brand starved is the opposite of the priority order the data suggests.
2. Recommend structural changes to Performance Max and branded search.
Read the worked answer
Sequence the changes so measurement stays interpretable.
- Apply brand exclusions to Performance Max where available, or use account-level negative keywords for brand terms so Performance Max stops harvesting brand queries. Expect reported return on Performance Max to fall; that drop is honesty, not deterioration. Brief finance before it happens.
- Keep branded search but right-size it. With 92% impression share, the goal is coverage and cost control, not growth. Check that match types and bids are not overpaying for a near-certain click, and consider a separate low-bid campaign with tight negatives for generic terms.
- Put incremental budget into non-brand. With 28% impression share and 40% lost to budget, this is where marginal dollars can buy new customers. Segment non-brand by category and margin before scaling, since not all revenue is equally profitable.
- Re-examine the 400% target on Performance Max. Actual delivery at 520% suggests the target may be throttling volume. Test a modest target reduction alongside a budget increase, and watch marginal return rather than average return.
Do not make all four changes at once; stage them over two to three weeks with a written read plan.
3. Design the test that would settle the branded search question for finance.
Read the worked answer
Finance needs an incrementality answer, not an attribution argument. Propose a geo-based brand pause test.
Design: select matched markets by historical branded revenue and seasonality, randomly assign matched pairs to a test group where branded search is paused for four to six weeks, and keep control markets live. Measure total revenue and total orders from all channels in both groups, not branded search conversions, because the question is whether paused clicks are recovered through organic results or lost to competitors.
Expected read: some branded clicks come back through organic listings, especially at position one, and some are lost. The net difference is the true value of the campaign. A typical result is that branded search is partially incremental, which supports keeping a right-sized campaign rather than eliminating it.
Guardrails: do not pause during peak season or a major promotion, and ensure the test markets are large enough to detect a small effect. Run the analysis at the total-business level and present it as a range with confidence bounds.
Close the loop by pairing this with the Performance Max brand exclusion work, so finance sees one coherent story: which spend creates demand and which spend defends it.