Search Filters That Lie: Facet Counts, Dead Ends, and the Fix Order
A shopper taps a filter, watches the count promise 14 products, and lands on 3. She does not think your data is messy. She thinks the store is small, or broken, and the next filter she tries is on a different site.
Facet counts are a trust surface. When the number beside a filter value disagrees with the results page, or a combination of two reasonable filters returns zero products with no explanation, the catalog itself takes the blame.
Where the counts go wrong
Counts usually break in one of three places. The index updates slower than inventory, so sold out variants still count. The count is calculated before another active filter is applied, so it describes a catalog the shopper is no longer looking at. Or variant level stock is rolled up to product level, promising sizes that do not exist. Each cause has a different fix, which is why guessing wastes a sprint.
Baymard Institute product list research treats filtering as a core buying tool, not decoration, and shoppers on larger catalogs rely on it to do the shortlisting a category page cannot. On a phone the problem compounds: filters hide behind a button, so a wrong count costs two taps and a reload to discover.
Zero results need a recovery, not a dead end
A zero result state should name which filter combination caused it, offer one tap removal of the most restrictive filter, and show the nearest matching products below. A blank grid with a reset link is the worst version: it makes the shopper do the diagnostic work.
Keep a log of zero result filter combinations for a month. Recurring combinations are merchandising information. If shoppers keep filtering for a size, material, or use case you do not stock, that is a buying signal or a labeling error, and either answer is useful. Related: How Edge Personalization Works Without Slowing the Storefront Down.
Fix order that respects effort
Fix count accuracy first, because every other filter decision depends on trusting the numbers. Fix the zero result page second, because it is one template and it rescues traffic you already have. Save filter redesign, sticky filter bars, and saved filter sets for last. They polish a system that may still be lying.
Shopify stores should check what their search and discovery setup recalculates in real time and what it batches. Shopify Help Center material on search and filtering is blunt about which behavior is native and which comes from an app, and that boundary is where most count bugs live. Audit there before paying for a redesign.
Frequently asked questions
Why do my filter counts not match the results?
The usual causes are a stale search index, counts calculated before other filters apply, and variant stock rolled up to product level. Reproduce one mismatch, note the exact filters used, and trace which of the three causes produces it.
Should I hide filters that return zero products?
Hide or disable them only if the count is accurate in real time. A greyed out filter with a zero count tells the shopper the truth. A clickable filter that lands on an empty page does not.
How many filters should a category page show?
Show the attributes shoppers use to shortlist in that category, usually a handful, and collapse the rest. Every filter you add is a combination that can dead end, so each one has to earn its place.
Do facet counts affect mobile shoppers more?
Yes. On mobile, filters sit behind a button and each check costs a reload. A wrong count that a desktop shopper notices in seconds can end a mobile session outright.