Brand search ROAS: Are you measuring new demand?

A strong account average can answer the wrong question

Suppose a team wants to grow acquisition and the report shows a 7x ROAS. Before using that number to justify more prospecting spend, ask where the revenue came from. An account can look efficient because much of its traffic already knows exactly which business it wants.

That does not make brand search useless. It means the account average cannot tell you how well the next dollar will reach someone new. Capturing demand and creating additional sales need different evidence.

Begin by stating the decision. Are you trying to protect access to existing demand, reach new customers or estimate how many sales would disappear without the ads? Choose the report or experiment that can answer that question.

Keep brand intent and customer status separate

A new customer can search your business name after a recommendation or another marketing interaction. A returning customer can click a non-brand query. Brand/non-brand classifies the search; new/returning classifies the customer relationship.

Use business records to define a new customer wherever possible. Decide how you handle repeat inquiries, changed email addresses and customers who bought a different service before. Keep uncertain classifications separate rather than forcing every record into a confident label.

For lead generation, add qualified opportunities and closed customers. A non-brand campaign that produces many forms may still fail to acquire suitable business. The report needs to preserve what happened after the click.

See how the revenue mix changes the headline result

Consider an illustrative account with $1,000 spent on brand search and $9,000 on non-brand search. Suppose the platform attributes $40,000 of revenue to brand and $30,000 to non-brand. The blended report shows $70,000 on $10,000 spend: 7x ROAS.

The components are very different: 40x for brand and about 3.33x for non-brand. Neither number includes delivery costs or proves incremental sales. The split simply reveals that the account average should not be used as a forecast for doubling the non-brand budget.

Now add customer status and margin. If the non-brand customers are valuable and mostly new, that segment may deserve investment despite its lower attributed ROAS. If the attributed brand sales would largely happen anyway, the high number may overstate the value of additional brand spend.

• Calculate acquisition economics beyond a headline ROAS

What the eBay experiments show and where they stop applying

Blake, Nosko and Tadelis studied paid search through large-scale field experiments at eBay. Their 2014 working paper reported no measurable short-term benefit from brand-keyword advertising in that setting and differences in non-brand effects across customer groups.

The finding matters because people who click ads can already have high purchase intent. A report can credit an ad for purchases that would have happened through another route. Observing the click and the sale does not establish the size of the causal contribution.

It would be a mistake to turn that study into “all businesses should stop brand ads.” eBay was a well-known marketplace, the study is historical and a local service business faces different competition and discovery conditions. Use the work to understand the measurement problem, not to import a universal budget rule.

• NBER: Consumer Heterogeneity and Paid Search Effectiveness

Choose an experiment that can detect a useful difference

A causal test needs a credible comparison with what would have happened without the advertising. Depending on scale and constraints, that might involve a randomized holdout or a carefully designed geographic test. The design should account for customer movement, spillover and sales timing.

A simple before-and-after pause is vulnerable to seasonality, competitor activity and other marketing changes. It can offer operational clues, but it should not automatically be presented as a causal estimate. Avoid changing the page, offer and sales process during the same comparison.

Small accounts may not have enough volume to resolve modest effects precisely. Decide the smallest difference that would change the budget decision and assess whether the design can measure it. If it cannot, report the uncertainty instead of using a weak test to produce a strong conclusion.

• Agree test boundaries and review rules before spending

Give each part of the account a stated job

A brand campaign might help people reach the intended destination or communicate a current offer. A prospecting campaign needs evidence that it can attract commercially suitable demand. Record those jobs before comparing their efficiency.

Keep the diagnostic report readable: spend, attributed outcomes, customer status, margin where available and the limitations of each measure. Do not label all non-brand revenue incremental or all brand revenue unnecessary.

Use the evidence to make a bounded decision. That might be a targeted experiment, a budget reallocation or a better customer-status report. The useful outcome is clarity about what the next dollar is supposed to accomplish.

Frequently asked questions