People who search buy more. That doesn't mean what it looks like

In short

• In any store, people who use the search convert several times more than people who don't. The difference is real and you can measure it at home.
• The usual conclusion ("getting more people to search increases sales") doesn't follow from it: the people who search already arrived with intent.
• Optimising for search usage leads to absurd decisions, such as hiding the navigation.
• The metrics that matter compare the search with itself over time, not with the rest of the site.

There's a number that turns up in almost every presentation of search tools: people who use internal search convert several times more than people who don't. The multiple varies with whoever quotes it, typically between two and six times, and the original source almost never comes with it. It's worth saying this straight away: it's a number we have no way of verifying, and this article doesn't depend on it.

What matters is that the direction of the effect holds in any store that measures the two groups. People who search convert more. Nobody has to believe us, or Econsultancy, for that: just open your own store's analytics.

The problem is the sentence that usually comes next: "so get more visitors using the search". That sentence doesn't follow from the number, and acting on it leads to decisions that hurt the store.

Why the correlation is real

There's no trick in the measurement. Someone typing in the search box is at a different stage of the visit. They've already decided they want something specific, they have a word in mind, and they spent effort writing it.

The visitor browsing the categories may be doing that, or may be strolling around, comparing prices, killing five minutes. The two groups are alike in nothing, and one of them was selected by a behaviour that is already, in itself, a signal of intent to buy.

This is called selection bias, and it's the simplest explanation for the number. The search didn't turn visitors into buyers: it's the buyers who use the search more.

What the correlation doesn't prove

It's worth separating three claims that usually travel glued together:

The second row is the one that matters, and it's the one that can be measured without ambiguity: among the people who search today, how many find and buy? If the search improves, that number goes up, and the effect is attributable. That's the sum the Optimisation screen does per term, when it estimates what each one recovers per month.

There's room for that second row almost everywhere, and it has been measured: in the 2026 revision of the Baymard Institute search benchmark, 56% of the sites assessed fail to meet visitors' search needs, and 58% of the mobile versions perform mediocrely or worse. The search that already exists is almost always the piece with the most slack to improve.

The third is the one you see put into practice and the one that does damage. We've seen stores hide the category menu to "increase search usage", and lose sales from people who were browsing for the pleasure of it.

The metrics that are worth it

The rule is to compare the search with itself, and not with the rest of the site. Four numbers are enough:

• Searches with no results, as a percentage of the total. It goes down when the catalogue's vocabulary gets closer to the customers'.
• Click rate on the results, that is, the fraction of searches that led to at least one click. Not clicks divided by searches: that sum goes past 100% the moment somebody clicks two products, and a rate that can read 130% isn't a rate.
• Searches per session among the people who search. If it goes up without sales going up, people are rephrasing because they can't find things, or fighting with a column of too many filters.
• Revenue attributed to the order lines that came from a click on the results. The tightest attribution there is, and the only honest one.

The last one deserves insisting on. Counting the whole order as search revenue because the person searched at some point is what makes these reports look miraculous. If the customer searched for socks and bought a coat, the search didn't sell the coat.

Pros and cons of optimising for search usage

How we look at this

The payback is measured by fixing the group of people who search and looking at what happens to that group before and after a change. A list of synonyms approved on a Monday has a visible effect on the empty results rate of that same week, and that effect is attributable because nothing else changed.

Revenue attribution follows the same principle of modesty. In Buskara's analytics screens, attributed revenue counts the order lines of the products that were clicked in the results, and not the whole order. The session that searched and bought without clicking shows up separately, as assisted, precisely because it's another thing. Two columns instead of one, and the bigger of the two isn't the one used to justify the tool.

There's a detail worth knowing when reading these numbers: the bridge between the search and the order is a cookie of our own, and in a store that requires consent it's only written once consent has been given. Orders from sessions that refused are not attributed. That's the correct behaviour, and it makes attributed revenue always a floor and never a ceiling.

It's all smaller than "six times the conversion", and it has the advantage of being true. It's also the sum worth doing before choosing a plan: the payback of a better search is measured in the group that already searches, and that group is already identified in today's reports.

Frequently asked questions

Is it true that people who use internal search convert more?

The difference shows up in practically every store that measures it, and each store can confirm it in its own analytics. The multiples going around the internet, between two and six times, are another conversation: they rarely come with a visible methodology. What the number doesn't say, whatever it happens to be, is the cause: the people who search already arrived intending to buy, and that intent explains a good part of the difference.

Should I push visitors towards the search box?

Not at the expense of the navigation. A visible search box with good suggestions is good practice; hiding categories to increase search usage is optimising the metric and hurting the business.

Which search metrics should I follow?

Searches with no results, the percentage of searches that led to at least one click, the number of searches per session among the people who search, and the revenue of the order lines that came from a click on the results.

How do you measure the payback of improving the search?

By comparing the same group (the people who search) before and after a concrete change, such as the approval of a list of synonyms. The comparison between people who search and people who don't is no use for this, because the two groups aren't comparable.

What is tight revenue attribution to the search?

It's counting only the order lines of the products the customer clicked in the search results, and not the whole order. It can give low values with many orders in the list, and that's information and not a defect of the measurement.