Site search & filtering once you pass 50 SKUs
Once your catalog passes roughly 50 SKUs, site search and filtering become revenue-critical, not a nicety. Visitors who search convert several times higher than those who only browse, so a failed search is lost money.
AI can cluster your search logs, spot the zero-result gaps, and propose synonyms and filters. You feed it the real queries and confirm which terms match which products.
Fastest path: one prompt, end to end
🤖 AI prompt — paste into ChatGPT / Claude
You are a site-search and merchandising analyst. Use MY data only.
Roughly how many SKUs + main product types: [e.g. 300 SKUs, apparel]
Attributes people shop by: [size / color / use case / material / price]
My top search queries + any zero-result queries: [paste, or say where to get them]
Do this:
1. From my zero-result and low-result queries, list likely SYNONYM gaps (buyer word vs my catalog word, e.g. "sneakers" vs "trainers") and the products each should map to.
2. Design a filter set from the attributes I listed: which filters, in what order, and which should be multi-select. Keep it to what shoppers actually use.
3. Tell me how to track zero-results going forward and what a healthy zero-result rate looks like as a direction.
4. Read my top queries as merchandising signal: what demand they reveal, what to feature or restock.
Do not invent my queries, SKU counts, or a zero-result rate. If I didn't paste real search data, tell me exactly where to export it (my search analytics) and stop, do not make numbers up.
Output: synonym map + filter design + zero-result tracking plan + merchandising reads.
Or do it in 4 steps
- Turn on and read zero-results. The highest-value report in search analytics is queries that returned nothing. Each one is a shopper who wanted something and hit a wall. Export them weekly. They tell you what to add, rename, or restock.
- Fix synonyms so real words find real products. Buyers use their words, not your catalog's: "sneakers" vs "trainers", "hoodie" vs "pullover", a brand name vs your SKU title. Map each real query term to the products it should surface, so no search dead-ends on vocabulary.
- Design filters around how people actually shop. Past 50 SKUs, browsing needs filters by the attributes that matter: size, color, use case, price, and material where relevant. Order them by what shoppers use most, and make the ones with many options (size, color) multi-select. Don't ship a filter nobody shops by.
- Read search as merchandising input. Your top queries are unpaid demand signal. They show what people expect you to carry, which products to feature on the homepage, and where a catalog gap is costing sales. Search analytics is the cheapest merchandising research you have.
Worked example (labeled): an apparel store at 300 SKUs. "waterproof jacket" is a top query with a high zero-result rate, because its catalog says "rain shell". Adding that synonym maps the query to 12 real products and recovers those searches.
The same logs show "petite" searched often with few results, a real catalog gap worth stocking. Filters get reordered to size, color, use case (the three shoppers actually use), with size and color multi-select. Conversion on searched sessions climbs, because searches now land instead of dead-ending.
Past 50 SKUs, treat search as revenue infrastructure: track zero-results, fix synonyms, filter by real attributes, and read the query log as free merchandising research.