Can't tell why your sales dropped? Sledge reads your store data and ranks the leaks by what fixing them is worth. Free for 14 days →

// Level 04 The Fixes · 4.12

Illustrated ruins in violet overgrown with roots, under a full moon.

Recovering dead-end searches

Updated August 2026 7 min read
SAME QUERY. SAME CATALOGUE. TWO PAGES. WHAT MOST STORES SHOW No results found. The visit ends here. WHAT KEEPS THE VISIT ALIVE No match for "navy runner". Did you mean: navy trainers? Closest in Footwear: Tell me when you stock it → And the query is now on a list. A FAILED SEARCH IS THE MOST SPECIFIC DEMAND SIGNAL YOU WILL EVER RECEIVE.

Nothing about the catalogue differs between the two panels. The recovery is a page and a list, not a purchasing decision.

The short answer

A zero-result search is two things at once, and most stores waste both. It is a visit about to end, recoverable with a page that offers close matches and a way to be told later instead of the words "no results found". And it is a ranked demand signal: a list of exactly what people came to your store expecting to find, in their own words, which is more specific than any keyword tool will ever give you. The first is a design job. The second is a purchasing and merchandising input, and it is free.

One event, two opportunities

Chapter 4.9 covered how to stop searches failing. This chapter is about the ones that fail anyway, because some always will. A failure is not only a loss.

Right now: a visit about to end

Compilations of search behaviour report that a large majority of shoppers who hit an unsuccessful search leave and buy elsewhere rather than trying again. They are on your site, at peak intent, and the store has just said no.

This is recoverable with a page, and the page costs nothing to build.

Afterwards: the most specific demand data you will ever hold

A keyword tool estimates what people search for on the internet. Your zero-result log records what people typed into your store, having already chosen to be there, in the words they actually use.

Ranked by frequency, that list is a purchasing brief, a synonym list and a product-title rewrite, all free.

A failed search is a customer telling you, unprompted and in their own words, exactly what they came for. Most stores answer it with four words and throw the rest away. The Dropshipping Playbook

What the page should do

"No results found" is the default in most themes and it is the worst available answer, because it terminates the interaction and offers no alternative except leaving.

A recovery page has four jobs, in order:

Acknowledge it
Repeat the query back. "No match for 'navy runner'" confirms the store understood, which a bare empty state does not.
Offer a correction
A spelling or synonym suggestion resolves the majority of these, because most zero results are vocabulary, not absence, per chapter 4.9.
Show something
The closest category, or your best sellers as a fallback. An imperfect grid beats an empty one, because it gives the visit somewhere to go.
Capture the intent
A notify-me for a genuine gap turns a lost visit into a contact address attached to a known want, which is worth more than a generic signup.

Note the third row's fallback. Showing your best sellers to someone who searched for something else is a weak answer, and still better than nothing, because it keeps the session alive long enough for the shopper to find a second reason to stay.

Triaging the list

Once you have a ranked log of failed queries, most of them fall into four buckets, and only one of them is a purchasing decision.

BucketWhat it meansWhat to doCost
You sell it, wrong wordVocabulary gapAdd a synonym, or retitle the productMinutes
You sell it, misspelledMatching too strictTurn on misspelling toleranceA setting
Not a product"returns", "shipping", "contact"Include content pages in searchA setting
You genuinely do not sell itA real gapNotify-me, then decide with the countA sourcing decision

Three of the four are free and take an afternoon. The fourth is the interesting one, and it is the only place in this playbook where a storefront signal turns into a supplier conversation: a query appearing hundreds of times a month, from people who chose to visit your store, is a stronger case for sourcing something than any keyword volume estimate.

The constraint, again

All of this depends on the queries existing somewhere, and by default they do not.

Chapter 3.4 made the general case: search queries are events rather than stored records, and nothing reconstructs them afterwards. A store that starts recording today has a triage list next month. A store that does not has the same blank page and no idea how often it is being shown.

Check the rate before the list

Published guidance puts a typical zero-result rate at 10 to 20% and best practice under 5%. If yours is not measured, it is not zero. Establish the rate first, because it tells you whether this chapter is a tidy-up or the largest single leak in your store.

In practice

  1. Replace the empty state with a page that acknowledges, corrects, shows something and captures. This helps every failed search from today, including ones you have not diagnosed.
  2. Start recording queries and their result counts, if nothing is.
  3. Each month, take the top ten failures and triage them into the four buckets above.
  4. Clear the three free buckets immediately. They are synonyms, settings and titles.
  5. Let the fourth accumulate a notify-me count before treating it as a sourcing decision.

Step one is worth doing first even with no data at all, because it is the only fix here that works on searches you never find out about.

Common questions

What should a zero-result page actually contain?
Acknowledge the query in the shopper's own words, offer a spelling correction if one is plausible, show the closest products or the nearest category rather than an empty grid, and give a way to be notified if you genuinely do not stock it. Anything is better than four words and a search box.
Should I stock what people search for and don't find?
Sometimes, and that is the point of keeping the list. A query appearing hundreds of times a month is a demand signal with a volume attached to it, unlike a keyword tool's estimate. But check first whether you already sell the thing under a different name, which is the more common case by far.
Is it worth offering a back-in-stock or notify-me option?
For genuine gaps, yes. It converts a lost visit into a contact address and, unlike a generic newsletter signup, you know precisely what that person wants. It also gives you a demand count for a product before you commit to sourcing it.
How do I know which failed searches cost the most?
Rank by frequency first and by the query's obvious commercial intent second. Something has to be recording them: search queries are events, and per chapter 3.4 they are not retained by default and cannot be recovered retroactively.
Illustrated violet rock outcrop with a yellow sun rising behind it.

The app behind these fixes

Most of Level 04 is a setting in Sledge rather than a project.

See it on the Shopify App Store