How B2B buyers actually search

B2B search is not discovery. In most of your traffic the buyer already knows what they want and is trying to type it in as fast as possible. Designing for that changes almost every decision you make about search.

Five patterns, and how much traffic each one is

Look at your own query log before you believe anyone's numbers, including these. But the shape is consistent across technischer Großhandel, Ersatzteilhandel and manufacturers selling direct:

The first three are all the same thing wearing different hats: identifier lookup. Together they are usually the largest share of B2B search traffic, and they are the easiest to get right — and the most expensive to get wrong, because a buyer who typed an exact number and got a relevance-ranked list of forty similar articles concludes your shop is broken. They are not wrong.

Exact-code match must beat fuzzy match

The rule is short to state and quietly broken: if the query matches an identifier exactly, that article is result one, and nothing outranks it — not a merchandising boost, not a campaign, not a higher-margin substitute.

What this requires in practice:

• Identifier fields are matched without typo tolerance. One allowed typo turns 4711-A into a match for 4711-B. Different article, different price, possibly a different thread.
• Identifiers are indexed so that partial entry from the front works — typing 4711 surfaces 4711-A in autocomplete — while the full string still wins when it is complete.
• Identifiers are also matched in the middle of a string, because buyers paste KAT-4711-A-2024 out of their own system and expect the article back.
• Normalisation is symmetric: if you strip spaces and hyphens when indexing, strip them the same way in the query. 6205 2RS, 6205-2RS and 62052RS are one article.

Manufacturer part number is a first-class field

In technischer Großhandel your own article number is often the least used identifier. The fitter knows the SKF number, the Festo number, the number stamped on the part. Your internal number means nothing to them.

Treat the manufacturer part number as a real attribute on the product — not as a line in the description, and not as a synonym rule. Reasons, in order of how much they will cost you if ignored:

• It has to be searchable and exactly matchable, which requires its own field with its own weight.
• It has to appear in exports and punchout catalogs, where a customer's procurement system matches on SUPPLIER_ALT_PID or the manufacturer's reference. A synonym rule lives in the search engine and never reaches a BMEcat export.
• It has to survive bulk maintenance, because manufacturers renumber. A field can be updated by import; a thousand synonym rules cannot.

The same argument applies to two more identifier fields worth modelling:

• Cross-reference numbers — the competitor's number for the equivalent article. Buyers type these constantly after a supplier change, and matching them is one of the highest-return things a B2B shop can do.
• Customer material numbers — the buyer's own number for your article. These are per-customer, so they are held on the customer relationship rather than on the product, and search resolves them for the signed-in organization. A buyer who can order by their number stops maintaining a translation list in Excel, which is the moment they stop shopping around.

Descriptive queries: short, technical, German

The word queries you get are not sentences. They are two to five tokens, mostly nouns, often mixing a material, a dimension and a norm: Winkel 90 Grad Edelstahl, Schlauchschelle 32 mm, Hutmutter M10 A2.

Three things follow:

• The dimension in the query is a filter, not a word. M10 should narrow to thread size M10 if you have that attribute, rather than compete as text. This is the pay-off for modelling technical facts as attributes instead of prose.
• Trade names and norm names are both live vocabulary. V4A, 1.4571, A4, Inox, NIRO and Edelstahl all get typed for the same material by different people in the same week.
• Compounds break word matching, as covered in How search works. Buyers type the head noun — Schraube, Schelle, Rohr — and your product names carry the compound.

The "I know the machine" case

The hardest pattern, and the one with the most revenue behind it: the buyer does not know the part. They know the machine, and that something on it broke.

Text search cannot solve this, and no amount of synonym work will. It is a data problem with three possible answers:

All three live in the configurator area, not in search — but search is where the buyer starts, so the two have to meet. A query like HK-40 should return the machine's parts page, not 200 loosely related articles. That is a redirect rule, and it is one of the highest-value rules you will ever write.

Zero results is your most valuable data

Every other analytics report tells you about people who succeeded. The zero-results report is the only one that tells you, by name, what buyers wanted and did not get. Each line is either a gap in your catalog, a gap in your vocabulary, or a genuine sales lead for something you do not stock.

It is also finite. Most B2B catalogs have a few dozen recurring zero-result queries and a long tail of typos. Working the top twenty once takes an afternoon and measurably moves the number of searches that end in an order. Working them every week keeps it moving. See Working the zero-results report.

Next

• Configure the index — make identifier fields behave the way this article demands.
• Autocomplete and suggestions — where identifier lookup actually pays off.