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 · 4.10

Illustrated rock spire standing in still violet water, framed by tree roots.

Filters that convert

Updated August 2026 7 min read
THREE CLICKS TO AN EMPTY PAGE Category picked 42 products + Size: M 9 products + Colour: navy 0 products WHAT THE SHOPPER SEES "No products match your selection." no route back except undoing their own choices NAVY WAS NEVER AVAILABLE IN M. THE INTERFACE OFFERED IT ANYWAY.

The counts are illustrative. What is not illustrative is the last step: most filter interfaces will happily offer an option that produces nothing.

The short answer

Filtering is the behaviour of a shopper doing exactly what you want: narrowing toward a purchase. The failure mode is that most filter interfaces offer every option regardless of whether it produces results, so a shopper picks size then colour and the page empties. They did nothing wrong. The interface offered a combination that never existed, and unlike a failed search, this dead end was built by the store rather than typed by the customer.

Filtering is a buying signal

Filters get treated as housekeeping.

A shopper applying a filter has a constraint. They need a size, a colour, a price ceiling, a material. That is not idle browsing; it is somebody with requirements narrowing toward a decision, which puts them close to the searchers in chapter 4.9 on the intent scale.

So when a filter interaction ends in an empty page, the visitor you lost was one of the good ones.

The dead end your interface built

Compare it with a failed search.

A search returns nothing because the shopper typed a word you did not anticipate. That is a vocabulary gap, and it is partly on them and mostly on you.

A filter returns nothing because your interface offered an option that produces no results and let them pick it. There is no vocabulary gap. The store presented navy as a choice, the shopper chose navy, and the store then reported that navy does not exist in the size they had already selected.

A failed search is a word you didn't anticipate. An empty filter page is a combination you offered. The Dropshipping Playbook

And unlike a search, there is usually no recovery path. A zero-result search page can at least suggest alternatives. An empty filtered collection typically says "no products match your selection" and leaves the shopper to reverse their own choices one at a time to work out which one was the problem.

Four rules that prevent almost all of it

Never offer a zero
Options that would produce no results should be disabled or removed as the shopper narrows. This single rule eliminates the dead end entirely.
Show the count
"Navy (0)" beside "Navy (7)" lets a shopper choose before committing a click. It costs nothing and it converts a dead end into information.
Always leave a way out
If the page does empty, offer "remove colour" as a one-click action, not just "clear all". Clearing everything discards work the shopper did correctly.
Keep filters in the URL
So a filtered view can be shared, bookmarked and returned to with the back button. A filter state that evaporates on navigation makes comparing anything impossible.

Your filter values are dirty

Before any of the above, most dropshipping stores have a data problem that makes the interface look amateur no matter how it behaves.

Supplier feeds arrive with inconsistent attribute values, and each variation becomes its own filter option:

What the filter showsWhat happenedWhat the shopper concludes
Blue, blue, Blue , BLUEFour casings and a trailing spaceNobody maintains this store
M, Medium, medThree suppliers, three conventionsWhich one is my size?
One value with 1 productAn attribute that is not really a filterPointless click
A filter with 40 valuesAn attribute with no useful groupingScrolls past it entirely
"Default Title"An untouched Shopify variant defaultThis is a template store

None of this is design work. It is an afternoon of data cleaning, and it is prerequisite: a beautifully behaved filter over dirty values still looks like a store nobody runs.

Which filters to offer at all

The right set is category-specific, and the test is simple: does this filter answer a question a shopper actually asks about this product?

Fewer filters also means fewer combinations, and the number of ways to empty a page grows with every filter you add. Each one you decline to offer is a set of dead ends that cannot happen.

You cannot audit this from the admin

Which combinations shoppers actually try, and which ones emptied, is an event stream rather than a report. Chapter 3.4 covers why none of it is retained by default and none of it is recoverable afterwards. Without it, you can fix the combinations you happen to guess and no others.

The order to do it in

  1. Clean the values. Merge casings, standardise sizes, delete "Default Title". Prerequisite for everything else.
  2. Cut filters that answer no real question. Fewer filters, fewer dead ends.
  3. Add counts to every option. The cheapest large improvement on the list.
  4. Disable zero-result options as the shopper narrows.
  5. Add a one-click way to remove the last filter, not just clear all.
  6. Start recording which combinations empty, so next month this is a ranked list rather than a guess.

Common questions

How many filters should a collection have?
As few as answer the questions shoppers actually ask about that category, and no filter whose values are not clean. Three well-populated filters beat eight where half the values return one product. Every extra filter also multiplies the number of combinations that can empty the page.
Should unavailable options be hidden or greyed out?
Greyed out is generally better than hidden, because a filter list that changes shape as you click is disorienting, and a visible unavailable option tells the shopper the store does stock that thing, just not in this combination. Either is far better than an option that looks available and yields nothing.
Why do my filters have weird values like 'Blue ' and 'blue'?
Supplier feeds. Imported product data arrives with inconsistent casing, trailing spaces and near-duplicate values, and each variation becomes its own filter option. This is the most common reason a filter list looks unprofessional, and it is data cleaning rather than design work.
Do filters matter on a small catalogue?
Less, and sometimes not at all. Under roughly fifty products a well-organised collection page can be scanned directly, and filters add a step rather than removing one. They start earning their place around the point described in chapter 3.5, where browsing stops working.
Illustrated violet forest at dusk, with a yellow sun between the trunks.

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