Interruption or Intent: Banners Against Search

Two Opposite Ways of Buying Attention

Display advertising buys attention from people who were doing something else. Search advertising buys attention from people who have just declared, in their own words, what they want. Every meaningful difference between the two channels — creative, targeting, response rate, measurement, and the argument in the budget meeting — follows from that single distinction.

The consequences run in opposite directions. Display can reach someone who does not yet know your category exists, which search cannot do, because nobody searches for a solution they have never heard of. Search can reach someone at the moment of active evaluation, which display can only approximate through inference.

That is why the standard comparison — which channel has the better return on ad spend — is the wrong question. One channel creates demand that does not yet exist in measurable form; the other harvests demand already expressed. Holding both to the same immediate-response standard guarantees that one looks like a bargain and the other looks like waste, and neither conclusion is trustworthy.

How the Banner Became the Default, Then Stopped Working

The banner was the web's first advertising format, and it arrived by analogy. Publishers had space, advertisers were used to buying space, and a rectangle on a page was a unit media buyers already knew how to purchase. Early response rates were, by every contemporary account, far higher than anything achievable now, because novelty was doing the work.

What followed is one of the better-documented declines in advertising. As inventory multiplied and readers learned the format, response rates fell steadily, and the industry answered by buying more impressions and adding animation, sound, and expanding formats. Each escalation bought back a little attention at the cost of some goodwill. Paid search meanwhile grew on a different model — pay for the click, tie the advertisement to a stated query — and took budget by being accountable in a way display was not.

Display did not disappear. It changed shape, moved into feeds and platforms, and is now consolidating into platform-managed campaigns: Google documents that Display Ads campaigns are moving into Demand Gen as the Display Network, with voluntary migration opening in June 2026. The rectangle survives. The expectation that anyone will click it does not.

Banner Blindness and the Ad Blocker

Two separate phenomena are routinely conflated, and they call for different responses. Banner blindness is behavioural: readers have learned where advertising sits on a page and do not look there. It is not hostility but efficient pattern recognition, and it defeats any format that depends on being noticed in a fixed position. Ad blocking is technical: software prevents the advertisement from loading at all, and it is common enough among technical audiences to matter for anyone selling to them.

The combination has a specific consequence in business markets. The audiences most likely to run blockers are frequently the exact audiences a technology company wants — engineers, system administrators, security teams. A display campaign aimed at that audience may be delivering to a systematically unrepresentative subset of it, and no platform report will disclose that.

The honest response is not to fight the format harder. It is to stop asking display for clicks. A creative approach built on the assumption that nobody will click, and that the value lies in the right person seeing the impression in the right context, at least matches how the medium behaves.

Where Display Still Earns Its Place, and Where It Does Not

Three uses hold up in business markets.

• Reaching a defined account list. When the target is two hundred named companies, search cannot reach them until someone there happens to search. Account-targeted display can put a company in front of a buying committee that has not started looking — the one job search structurally cannot do.
• Sustaining awareness where replacement cycles are long. If the equipment or the contract turns over every seven years, the buying window opens rarely and unpredictably, and being remembered when it opens has value no click-based measure captures.
• Supporting a launch. A genuinely new category has no search volume because the words do not exist yet. Display, trade media and public relations establish the vocabulary; search then harvests what they created.

Three uses do not hold up: broad untargeted display bought on reach by a company whose addressable market is a few thousand people, where nearly every impression is waste by definition; display bought to generate leads directly and judged on cost per lead, which produces either nothing or accidental clicks; and display bought because the budget had to be spent and the reporting looked busy. If a display line cannot be tied to one of the three defensible jobs, the money belongs elsewhere.

The Measurement Asymmetry Nobody Corrects

Comparing the reported return on ad spend of a display campaign against a search campaign is meaningless. The two figures come from measurement systems with opposite biases.

Display platforms credit themselves for view-through conversions: someone was served an impression, did not click, converted later, and the impression takes credit. Search sits at the end of the journey and takes credit under a last-click model, still the default frame in most internal reporting. The same conversion can therefore be claimed by display on a view-through basis and by search on a last-click basis, with both reports internally consistent while describing one deal.

The asymmetry compounds where cycles run across quarters. A display impression served in the first quarter and a branded search click in the third both precede a deal closing in the fourth. Which caused it is not knowable from platform reporting, because platforms are not measuring causation; they are counting coincidences inside their own attribution windows. The only reliable answer is experimental: hold out a matched set of accounts or a geography from the display buy and compare outcomes against the exposed set.

Programmatic and the Transparency Problem

Programmatic buying — automated, auction-based purchasing of impressions across many publishers through intermediaries — solved a real problem and created a new one. It made precise targeting and enormous scale available without negotiating with individual publishers. It also inserted a chain of intermediaries between the money and the impression, each taking a fee, and made it genuinely difficult for an advertiser to state where an advertisement appeared.

Three issues deserve a buyer's attention. The share of spend reaching the publisher rather than the supply chain is invisible in most reporting. Inventory quality is uneven: sites built primarily to carry advertising exist to absorb programmatic budget, and they will take a business advertiser's money happily. And viewability — whether an impression was ever on screen — is measured by standards the advertiser is buying against but not auditing.

None of this makes programmatic unusable. It makes it a channel that requires the buyer to demand a site-level placement report, read it, exclude aggressively, and insist on knowing the fees.

Retargeting: The Exception That Confirms the Rule

Retargeting is display bought against intent, which is exactly why it outperforms the rest of display and exactly why its reporting is the most flattering in the channel. The audience is defined by a behaviour — a site visit — rather than an inferred attribute, so the format finally has something in common with search.

It is also the easiest place in digital advertising to fool yourself. A retargeting campaign shows advertisements to people who already visited the site, some of whom were coming back anyway. Any conversion those people complete is available to be claimed by the campaign. The reported cost per acquisition looks better than every other line, and part of it is bookkeeping rather than incremental sales.

The same experimental discipline settles it: withhold retargeting from a randomly chosen share of the eligible audience and compare. Companies that run that test are frequently surprised in both directions — some find the channel is doing almost nothing, others find it is genuinely holding a long consideration process together.

They Answer Different Questions

The conclusion is not that search wins. It is that the two channels answer questions a business needs answered at different moments, and that a plan containing only one of them has a predictable blind spot.

Search answers: when someone in our market decides to look for this, are we there, and can we convert them? A company that cannot answer yes has an urgent and comparatively cheap problem to fix before any display money is spent. Display answers: are the people who will one day run that search aware of us, and did we reach the accounts that matter before they started looking? A company with strong search coverage and a shrinking pool of searchers has a demand problem that more bidding cannot solve.

So fund search coverage for every commercial and transactional query in your category first, because that demand exists and someone will capture it. Then fund display against a named target and a named job — account coverage, launch support, or presence across a long replacement cycle — with a holdout group attached from the first day. And refuse, politely and permanently, to compare the two channels on the same reported return figure.

Frequently Asked Questions