Testing Search Bar Placement: Icon, Field, or Both

Test three header states where feasible: icon only, a visible search field, and icon plus a search forward menu or home section. Judge the test on search usage rate, the share of sessions that search at all, and on revenue per searcher, not raw conversion, because placement changes who searches as much as how well search works.

The header argument is always framed as space. A visible field takes room the logo and the sale banner want. So search retires behind a magnifying glass icon, the design breathes, and nobody checks what happened to the shoppers who used to type.

Searchers are not average visitors. On most catalogs they convert at a multiple of browsers because they arrived with a named want. Any design that quietly reduces the number of people who search is taxing your highest intent traffic to pay for whitespace.

Define the metrics before the variants

Primary metric: search usage rate, sessions with at least one search divided by all sessions, tracked separately for mobile and desktop and for new versus returning visitors. Secondary: revenue per searcher and the zero results rate, which tells you whether new searchers brought worse queries or the same queries to a weaker results page.

Run placement long enough to cover weekday and weekend mixes, and beware the novelty window. A newly visible field attracts curious clicks in week one. Read the result from the later weeks, when the design has stopped being news.

What the variants usually reveal

On deep catalogs, a visible field tends to lift usage among new visitors most, because returning visitors have learned the icon. On small curated catalogs, usage barely moves and the icon is a fair choice. The catalog decides, which is why imported advice from a store with a different assortment keeps misleading people.

Also test the middle option honestly: icon in the header, plus a real search field at the top of the menu drawer and on the zero state of the collection page. Some stores recover most of the usage with none of the permanent header cost, but you only learn that by measuring the drawer variant as its own arm, not as an afterthought.

Guardrails while you test

Hold the search engine and results ranking constant across variants, or you will attribute an algorithm change to a placement change. Freeze merchandising pushes that mention search during the test window. And check that the visible field does not slow the header on mobile; an input that delays interactivity trades one tax for another.

Decide the draw rule in advance. If usage rises but revenue per searcher falls by the same proportion, total search revenue is flat and the prettier header wins. Write that rule down before the data arrives, or the review meeting will invent a metric that flatters the preferred design.

Frequently asked questions

What is the main metric for a search placement test?

Search usage rate, the share of sessions that perform at least one search, paired with revenue per searcher. Placement changes how many people search, so judging only the conversion of those who did search misses the effect.

Do shoppers convert more when they use search?

Typically yes, by a wide margin, because search signals a specific intent. That is why placement matters: designs that suppress search usage shrink your highest intent segment even if the results page itself is unchanged.

Should mobile and desktop search placement match?

Not necessarily. Screen space and behavior differ, and tests often show a visible field paying off more on desktop while a prominent icon plus menu field suffices on mobile. Split the analysis by device and decide separately.

How long should a header search test run?

At least two full weekly cycles after a short novelty window, longer for low traffic stores. Placement effects are behavioral and need weekday and weekend traffic to settle into a trustworthy read.