The Zero Results Audit: Fixing On-Site Search That Finds Nothing
A clear order confirmation email answers the questions buyers ask support most often: what was bought, what it cost, when it ships, and how to track it. Getting those details right cuts ticket volume in the busiest week of the order cycle.
A clear order confirmation email answers the questions buyers ask support most often: what was bought, what it cost, when it ships, and how to track it. Getting those details right cuts ticket volume in the busiest week of the order cycle.
This guide walks through order confirmation emails in plain steps, with the trade offs stated honestly and a short list of checks you can run yourself. Nothing here needs special equipment or a professional call out to get started, and each section ends with the decision that actually matters.
Address validation checks a typed address against postal data as the shopper types, then suggests the corrected version. Used well it cuts failed deliveries and reshipping costs. Used badly it rejects valid rural, military, and new build addresses, so always let the shopper keep their original entry.
A good checkout error names the exact field, says what went wrong in plain words, and tells the shopper the one action that fixes it. Errors should appear next to the field, keep everything the shopper already typed, and never blame the shopper for a formatting difference the store could handle itself.
Neither format wins on its own. Drawers suit small baskets and single product stores because they keep the shopper on the page they were browsing. Full cart pages suit larger baskets, discount codes, and gift options, where shoppers want room to review before paying. Pick based on how your shoppers actually edit a basket, not on what looks modern.
Pull thirty days of internal search terms that returned nothing and group them into misspellings, synonyms your catalogue does not use, products you do not stock, and formatting traps like plurals and hyphenated names. The first, second, and fourth groups are fixable this week with redirects and synonym rules. The third group is merchandising intelligence you are currently throwing away.
Four kinds of nothing
Zero results pages look identical to the shopper, but the causes split four ways and each has a different fix. Misspellings need tolerant matching, not a lecture. Synonym gaps, where the shopper types trainers and your catalogue says sneakers, need a synonym table. Out of stock and never stocked terms need a decision. Formatting traps, a plural or a model number typed with spaces, need normalisation rules most search apps already have switched off by default.
Count each group before touching settings. In our experience auditing stores, synonym gaps and formatting traps usually outweigh true misspellings, which matters because they are also the cheapest to fix. If your search tool cannot export the zero results list, that is the first problem to solve. Flying blind here costs more than any single fix.
Synonyms are catalogue work, not search work
A synonym table built in one sitting from one person's vocabulary will miss how shoppers actually talk. Build it from the zero results report itself, month by month, and add the regional words too: the same store can get jumper, sweater, and pullover for one rack of knitwear. Each entry should point at the collection or the attribute shoppers mean, not at a single product that will sell out and strand the rule.
Review the table when ranges change. Seasonal synonyms rot: a rule written for last winter's boots points at a page you unpublished in spring, and the shopper lands somewhere worse than a clean no results page. Treat the table like the catalogue it describes, with an owner and a review date.
When nothing is the honest answer
Some zero results searches are for products you do not sell and should not pretend to. The failure is a dead end page that says no results and offers nothing. A useful empty state names the closest category, shows the best selling items in it, and offers a notify me or contact route for the thing they actually wanted. That page converts a small share of disappointed shoppers and, just as usefully, logs demand.
Track those demand terms separately from the fixable groups. A term searched two hundred times a quarter for a product adjacent to your range is a buying signal with a number attached. It belongs in a range review, quoted with its search count, not in an anecdote about what customers keep asking for.
Measure the fix like an operator
After the changes, watch two numbers for a month: the share of searches returning zero results, and revenue per search session split by shoppers who saw a zero results page versus those who did not. The first number should fall. The second tells you whether the redirects land shoppers somewhere they will actually buy from, or just somewhere that technically exists.
Do not judge it in the first week. Search behaviour shifts with campaigns and payday cycles, and a big email drop can flood the report with one product name typed fifty ways. Give it a full trading month, compare against the same weeks, and keep the before export. Without the baseline you will be arguing from impressions by Christmas.
Keep reading: How Edge Personalization Works Without Slowing the Storefront Down, Edge Observability: Seeing Storefront Errors Where They Actually Happen, and Image and Font Delivery at the Edge: The Quiet Speed Win cover the nearby cases in this series.
Frequently asked questions
Where do I find zero results searches?
Most search apps and analytics tools keep a site search report listing terms with no results. Export thirty days, group the terms by cause, and fix the largest groups first rather than working the list top to bottom.
Should every zero results term get a redirect?
No. Misspellings, synonyms, and formatting variants should resolve to real results. Terms for products you do not stock belong on a useful empty state page that logs the demand instead of faking a match.
How big should a synonym table get?
As big as your shoppers' vocabulary, built gradually from the actual zero results report. A table grown from real searches beats a large one guessed in advance, and it needs a review whenever ranges change.
How do I know the audit worked?
Compare the share of searches returning zero results and revenue per search session before and after, over a full trading month. A falling zero rate with flat revenue means the redirects land in the wrong places.