Should You Bid on Your Own Brand? How to Measure Brand Search Incrementality

Brand search often looks like one of the best investments in a paid-search account. Conversion rates are high, cost per acquisition is low and attributed ROAS can be considerably better than almost anything else in the media plan.

That performance is not particularly surprising. Someone typing a company's name into Google is already demonstrating an unusually high level of intent. In many cases they know where they want to go and are using search as navigation. The advertisement appears immediately before the conversion, which makes the relationship between the click and the eventual purchase extremely easy for an attribution system to observe.

The commercial question is whether the advertising materially changed what happened next.

If a customer searches your brand, clicks the paid result and purchases, paid search can correctly record that interaction. What the observation cannot establish is whether the customer would simply have clicked the organic result, navigated directly to the website or purchased through another route if the advertisement had not appeared.

This is why I do not think the debate around bidding on your own brand should be reduced to a universal yes or no. There are businesses where branded paid search protects meaningful revenue, and others where a substantial proportion of the spend is buying access to customers who were already trying to reach the company.

The answer depends on the competitive environment, the level of organic substitution and, crucially, how the incremental return changes as more money is invested.

A high attributed ROAS cannot answer those questions.

Why brand search produces such attractive attribution

Paid brand search occupies an unusual position in the customer journey because the query itself tells us that demand already exists.

A customer cannot search for your company by name without first knowing the name. Something happened before that search to create awareness and, potentially, preference. It might have been previous experience, television, paid social, a recommendation, a physical store, a creator, word of mouth or simply years of exposure to the brand.

Paid search therefore frequently appears after much of the work responsible for creating the customer's interest has already occurred.

That does not make the search advert worthless. It means the revenue reported after the click needs to be interpreted in the context of the demand that generated the query.

Imagine a company spends £100,000 on its own brand terms and receives £2 million of attributed revenue. An attributed ROAS of 20 looks exceptional, but it tells us surprisingly little about whether the company should continue spending the full £100,000. If nearly all of those customers would have purchased through organic search anyway, the incremental economics will be very different from a situation where removing paid coverage causes a meaningful number of customers to choose a competitor.

Both businesses can show a brand-search ROAS of 20.

What matters is what changes when the advertising changes.

This is one reason experimental evidence on branded search has produced very different results between companies. The well-known eBay experiments found almost no short-term incremental benefit from advertising against its own brand terms, while later experimentation at Edmunds found a substantial proportion of paid branded traffic was not recovered through alternative routes when the advertising disappeared.

Rather than proving that one of those conclusions is generally correct, the difference illustrates the actual problem. Brand search incrementality depends on the environment in which the brand operates.

Organic substitution is the first thing to understand

The most obvious source of non-incrementality is substitution between paid and organic search.

Suppose 100,000 people search a company's name every month. With paid advertising running, 60,000 click the advertisement and 30,000 click the organic result. Paid search then reports the revenue associated with the 60,000 visits.

If the advertisement disappears, those people do not necessarily disappear.

Some may move to the organic result. Some may navigate directly to the website. Some may click another search result. Some may abandon the journey, and some may choose a competitor.

The economics of bidding on the brand are determined by that redistribution rather than by the number of conversions attributed to the paid listing while it is active.

This is why I like looking at paid brand and branded organic behaviour together before doing anything more sophisticated. If paid brand spend rises while paid clicks increase and branded organic clicks fall, there may be substitution taking place. If paid investment is reduced and much of the lost paid traffic subsequently appears in organic search while total revenue remains stable, the hypothesis becomes stronger.

This observed relationship does not prove incrementality. Paid and organic behaviour can move for other reasons and the customers selecting the two results have not been randomly assigned.

It does, however, tell us whether there is enough evidence of cannibalisation to justify asking a more demanding causal question.

For the same reason, I would not use paid brand search as evidence that underlying demand is increasing. Branded organic demand tells us much more about whether people are actively looking for the company. The decision to pay Google for access to that demand is a separate measurement problem.

Competition can make paid brand search genuinely incremental

The substitution argument changes when competitors are actively bidding on the brand.

If the organic result dominates the page and no meaningful competitor advertising appears around it, removing paid coverage may mostly shift traffic from a paid link to an organic one. When competitors consistently appear above or alongside the company's own results, the consequences can be very different.

The competitor is not necessarily creating the original demand either. They are attempting to intercept demand already associated with another brand. But from the perspective of the company being searched, preventing that interception can have genuine economic value.

I saw this directly at Oracle. In a highly competitive environment, we found incrementality in branded terms. That is one reason I have never found the blanket recommendation to stop bidding on your own brand particularly convincing.

The important difference is the market context.

Where competitors aggressively bid on brand queries, paid coverage can function partly as demand protection. Where there is little competitive pressure and the company has dominant organic visibility, a much larger proportion of the paid traffic may simply substitute for clicks the company would have received for free.

This also means competitor activity should not be treated as a footnote in the analysis. Auction behaviour is an effect modifier. The incremental value of exactly the same brand-search campaign can change when competitors enter or leave the auction.

Before altering the investment, I therefore want to understand who is bidding, how frequently they appear, how often they outrank the company's advertisements and whether the competitive pressure is concentrated around a small group of commercially important queries or spread across the entire brand account.

Those observations still do not give us an incremental ROI, but they tell us what kind of environment we are testing.

Incrementality is not a yes-or-no property of brand search

There is another reason I dislike the question "Is brand search incremental?"

It implies that incrementality is a fixed property of the campaign.

In reality, the incremental return can change with the amount being spent.

The first portion of brand-search investment might be highly productive. It may protect the highest-value queries, maintain visibility when competitors are present and prevent genuinely valuable customers from being intercepted.

As more money is added, the campaign can progressively buy additional coverage where the likelihood of losing the customer without the advertisement is lower. Bids rise, impression share approaches its practical ceiling and more of the additional paid clicks may come from customers who would have selected the organic result anyway.

The channel remains incremental.

The final portion of spend may not be.

This is simply saturation expressed inside brand search.

Imagine the first £200,000 of investment protects a meaningful amount of profitable revenue. Another £100,000 still creates value but at a lower return. The final £100,000 mainly increases paid coverage among searches where customers would almost certainly have reached the company through another route.

Looking at the average incremental return across the full £400,000 can hide that deterioration.

The allocation question is therefore not only whether paid brand advertising has some incremental effect. It is how much of the current investment remains incremental at the margin.

This matters because finding incrementality is sometimes treated as permission to maintain or scale the current spend. It should not be.

A campaign can be incremental and still be overfunded.

Why attributed ROAS tells us very little about the optimal spend

The saturation problem is almost invisible if the account is managed primarily through attributed ROAS.

Suppose underlying brand demand increases substantially. More customers search for the company, which creates more auctions, more paid clicks and more conversions. Paid-search revenue rises and the channel appears to be performing exceptionally well.

But the demand may have been created somewhere else.

The company may have run better creative. Television investment may have increased. Distribution may have expanded. Product reputation may have improved. More customers may simply have become willing to buy from the brand.

Paid search benefits because there is now more demand available to capture.

This creates the same confounding problem that makes search difficult to interpret inside an MMM. Search volume is partly an outcome of underlying demand, while paid-search activity also increases as that demand rises. Without accounting for the process creating the queries, it is easy to attribute demand creation to the channel capturing it.

Brand search makes this particularly easy to understand because a branded query is itself an expression of existing awareness.

That is why an extraordinarily high attributed brand-search ROAS does not automatically make me want to increase the budget. I want to know whether additional spend continues to alter the business outcome or simply buys a larger proportion of the traffic already trying to reach the company.

The difference between average and marginal return matters here just as much as it does in any other marketing channel.

Measure the business outcome, not the paid-search conversion count

Where the amount of money involved justifies stronger causal evidence, the next step is to create variation deliberately.

For sufficiently large businesses, this can involve reducing or removing brand-search coverage in selected comparable geographies while maintaining the existing investment elsewhere, then comparing the resulting business outcomes.

The outcome used to judge the test matters.

Paid-search clicks and paid-search conversions will obviously fall when paid-search advertising is reduced. That is not evidence that the business lost those customers.

If paid conversions decline by £1 million while branded organic revenue increases by £800,000 and total revenue barely changes, the economic loss is very different from the attribution loss.

Organic substitution is therefore part of the result rather than contamination that should be removed from the analysis.

The estimand should match the business decision. If we want to know whether £2 million of brand-search spend is necessary, we need to measure what happens to the business when that spend changes, including the routes customers use instead.

This is different from attempting to estimate some abstract causal effect of a paid click with every other route held artificially constant. The real-world business question is whether paying for the advertisement changes enough profitable customer behaviour to justify the cost.

A single on-off test may still not tell you the optimal budget

A full holdout can establish whether the current level of brand-search investment creates meaningful incremental value. It does not necessarily tell us whether the business needs to spend the full amount to capture that value.

Suppose a company spends £1 million on branded search and removing the entire campaign causes a meaningful revenue decline.

We have established that at least some of the investment is incremental.

We have not established that £1 million is the optimal spend.

Perhaps £400,000 of carefully targeted investment protects most of the revenue at risk. The remaining £600,000 could be operating on the flatter part of the response curve.

Where the size of the programme and available markets make it practical, varying treatment intensity can provide a much more useful answer than a single on-off comparison. Different reductions in spend, bidding intensity or coverage can help establish how the incremental response changes as investment falls.

That does not mean every company needs an elaborate multi-cell geo experiment for brand search. Often the amount at stake does not justify it, and observed substitution patterns combined with smaller controlled changes may be enough to make a sensible decision.

The complexity of the measurement should match the size of the allocation question.

But conceptually the goal remains the same: estimate how much incremental value is being created at different levels of investment rather than attach one permanent label to the channel.

Geo experiments have their own limitations

Brand-search experiments inherit the usual limitations of geographic testing.

Competitive pressure can vary between regions. Brand strength can differ. Customer behaviour, organic visibility and the proportion of revenue flowing through search may not be identical across markets.

Selecting treatment and control markets because historical revenue happens to look similar does not establish that they will respond similarly when paid brand coverage changes.

If competition is one of the mechanisms that makes the advertising incremental, I particularly want to understand whether the competitive auction environment in the tested markets resembles the environment where the result will eventually be applied.

The treatment itself also needs to be verified. If supposed holdout regions continue receiving a meaningful amount of brand advertising because of location settings or platform delivery, the contrast between the groups is weaker than planned.

And once an effect is identified, transportability remains a separate question. Finding an incremental return in a group of regions does not guarantee an identical return nationally if brand strength, competition and substitution behaviour vary across the country.

This is why incrementality is more nuanced than asking whether revenue would have happened anyway. The useful questions concern the size of the effect, the treatment actually delivered, the conditions under which the response occurred, the uncertainty around the estimate and how far the result can reasonably support the wider allocation decision.

The optimal answer may be less spend rather than zero spend

The practical consequence of all this is that the decision rarely needs to be ideological.

There are cases where stopping brand search almost entirely makes sense. There are others where aggressive competitive bidding makes maintaining strong coverage economically rational.

There is a large middle ground between those positions.

A company might discover that brand search is valuable on certain high-intent queries while a large amount of additional spend produces very little incremental benefit. It might maintain defensive coverage where competitors are most active while reducing investment where organic substitution is extremely high.

Another company might find that incremental returns remain attractive across most of the current spend but begin deteriorating as impression share approaches its upper range. The right decision would then be to hold rather than scale.

This is a much more useful outcome than declaring brand search either good or bad.

The objective is to locate the point on the response curve where the marginal incremental return no longer justifies keeping another pound in the channel.

That return should then be compared with the opportunities available elsewhere in the marketing mix.

Brand search should not receive credit for creating the demand behind the query

There is a broader allocation problem underneath brand-search measurement.

Suppose a company invests in advertising that increases awareness and preference. More people start considering the company. Branded search volume rises as a result.

Those customers eventually search the company name, click paid advertisements and purchase.

If the organisation evaluates marketing entirely through attribution, paid search can appear to be producing exceptional performance precisely because other marketing has successfully created more people willing to search for the brand.

The activity responsible for creating preference may struggle to demonstrate the same immediate ROAS because the eventual conversion is claimed by the channel closest to the transaction.

If budgets are allocated entirely according to those attributed numbers, capital progressively moves towards harvesting existing demand rather than creating more of it.

That is the broader reason the incremental economics of brand search matter.

The advertisement can provide genuine economic value without deserving credit for the preference that caused the search.

A business needs to distinguish those things if it wants to allocate marketing capital intelligently.

How I would evaluate brand-search investment

I would start with the observable relationship between paid and organic brand search.

Separate brand and non-brand activity properly, then examine branded organic clicks, paid clicks, spend, impressions and revenue over time. Look for periods where paid investment or coverage changed and examine whether organic behaviour moved in the opposite direction.

Then understand the competitive environment. Who is appearing against the brand? How frequently? Are they consistently appearing above the company's own listings? Does the competitive pressure vary materially by query or geography?

Those two layers usually tell us whether the current spend deserves deeper investigation.

The next question is materiality. There is little reason to design a complicated experiment to optimise a tiny brand-search budget. If the company is spending millions on branded queries, the potential value of distinguishing genuinely defensive spend from substitutable spend becomes much larger.

Where stronger causal evidence is justified, create deliberate variation and measure total business outcomes rather than the conversions attributed to the paid-search account. Follow substitution into organic and direct routes because that redistribution is central to the economics.

If possible and commercially worthwhile, test more than one level of investment rather than treating the question as purely on or off. The important output is the marginal response to additional spend.

Finally, keep monitoring the environment. A result produced when competitor pressure is low may no longer describe the economics if aggressive competitors enter the auction later.

The appropriate brand-search policy should therefore be evidence-based and adaptable rather than permanent.

There is no universal answer to whether you should bid on your own brand

The fact that major experiments have produced different brand-search results should make us suspicious of universal recommendations.

eBay found almost no short-term incremental benefit from its branded paid-search advertising. Edmunds found substantial traffic loss when branded paid links were removed. At Oracle, we found incrementality in a highly competitive brand environment.

Those findings do not contradict each other simply because they are different.

They describe different businesses operating in different competitive environments with different levels of organic substitution.

The more interesting question is what happens within each business as investment changes.

A brand can have genuinely incremental paid-search activity while still spending too much on it. The first part of the budget can protect valuable demand while later spend moves increasingly towards customers who would have arrived anyway.

That is why I would not ask simply whether brand search is incremental.

I would ask how incremental it is at the current level of spend, how quickly the return is saturating, what competitive pressure the investment is protecting against and whether the next pound could create more value somewhere else.

Growth Dynamics evaluates brand search using observable paid and organic behaviour first, followed by designed causal evidence when the size of the allocation decision justifies it. We look at substitution, competitive pressure, treatment intensity and marginal return rather than relying on attributed ROAS or applying a universal rule about whether branded bidding is good or bad.

The objective is not to prove that brand search works.

It is to determine how much of it is worth paying for.