Too many filters are a place where the sale dies

In short

• A column with twenty filters doesn't give more control: it gives more work before the first product is seen.
• A filter is only worth it if it genuinely splits the set on screen. With nine results, almost none of them do.
• Order matters more than quantity: price, category, and the two or three characteristics that catalogue lives on.
• Filters with a single value, or that would leave the list empty, should disappear on their own.

There's a way to make a store look well equipped without improving anything: show every possible filter. Brand, colour, size, material, price range, rating, availability, country of origin, collection, fastening type. It's impressive in a screenshot and it's terrible for someone looking for a coat.

The reason is easy to describe and easy to forget: every filter on screen is a decision demanded of the visitor before they see any products. Twenty filters are twenty decisions, and the person came to buy, not to configure a query.

What happens with twenty facets

On narrow screens, the filter column either pushes the products down or collapses into a panel nobody opens. In both cases the result is the same: the list stops being visible and the visitor goes back.

The difference a well resolved list makes is big enough to have been measured. In the Baymard Institute research on product lists and filtering, based on more than 4400 test sessions and a benchmark of 344 stores, sites with mediocre list usability recorded abandonment rates of 67% to 90%, against 17% to 33% among those whose tools were at least minimally tuned.

On a wide screen there's room, but another problem shows up: filters that split nothing. A search that returned nine results shows a brand filter with seven brands, a colour filter with eight colours and a size filter with eleven sizes. None of them helps anyone decide, and all of them take up the attention that should be on the nine products.

When a filter earns its place

The rule we use has three conditions, and the filter has to meet all three:

1. It splits the current set. If one value covers more than 90% of the results, the filter separates nothing. If each value has one result, it doesn't either.
2. It's a characteristic people decide by in that catalogue. In a shoe store, size decides. In a bookshop, it doesn't exist. If the characteristic isn't in the catalogue but people decide by it, the problem comes before the filters and it's covered in your catalogue is written for the supplier.
3. Its values are ones buyers recognise. "Material: PU" isn't a filter, it's a code.

Applied seriously, these three conditions usually leave between four and seven filters, which is the amount that fits on a screen without hiding the list. That's why, at Buskara, the column is built from the results of each search and not from a fixed list per category.

How many filters to show, by type of result

Notice that the last column never reaches twenty. In a large catalogue, the answer isn't to show more filters: it's to choose better which ones to show and leave the others one click away.

The counter is what makes a filter honest

There's one detail that changes how the column gets used more than the choice of filters does: showing, next to each value, how many products are left if it's picked.

Without a counter, every click is a bet. The person picks "red", waits for the load and finds out there are two products. With a counter, they see the two before clicking and pick something else. It's the same information, delivered before the decision instead of after, and that's what avoids the run of clicks that ends in "no results" and in leaving the site.

The counter has another, less obvious effect: it makes the catalogue's problem visible. A characteristic where 95% of the products have the same value shows up straight away in the numbers, and it becomes obvious that this filter splits nothing and is taking the place of one that did.

Pros and cons of showing everything

The fourth row of the left column is real and shouldn't be dismissed: for a share of visitors, a rich column signals a deep catalogue. The question is whether that offsets the cost for everyone else, and the cost is paid on every visit while the signal is only read once.

Rules we apply by default

• A filter with a single value doesn't appear. As a filter, it would show what's already on screen.
• Values that would leave the list empty don't appear. A filter that only leads to "no results" is a trap.
• The column comes from the search and not from the category. A search whose results have colour and size gets those two boxes; one where no result has a colour doesn't show it. It isn't a per category setting, it's a consequence of the set that's on screen.
• A box can start closed instead of open. It doesn't hide the filter, it shortens the column, and it's what saves categories with forty values.
• Whatever the visitor opens or closes beats the store's choice and lasts for their visit.

None of these rules is an aesthetic opinion. They all translate into fewer steps between the search and the product, and that's the number that counts.

Which of the catalogue's characteristics get to be filters is the merchant's decision, made once in catalogue filters, with a limit of twelve and a rehearsal alongside that shows, before saving, which boxes would appear for a term typed as a test. In PrestaShop, the characteristics and the attributes of the combinations are the raw material for this, with no extra catalogue work: see what changes with the search module.

Frequently asked questions

How many filters should an online store search have?

Between four and seven in the visible column, with the rest closed. Below ten results, the column can disappear entirely: choosing between ten products is faster by looking than by filtering.

Do too many filters hurt sales?

They hurt when they take up the product list's space or ask for decisions before anything is seen. The effect is bigger on a phone, where the column pushes the results off the screen or stays hidden in a panel.

Which filters should come first?

Price and category almost always, followed by the characteristic people decide by in that catalogue (size in footwear, power in tools, author in a bookshop). The order after that should come from real usage, not from the catalogue's structure.

Should I hide filters that return no results?

Yes. A filter value that would lead to an empty list should disappear, and so should a filter with a single value. In both cases, what it offers is work with no consequence.

How do I know which filters my customers use?

From the facet usage log: which ones get clicked, in what order, and which ones always appear without ever being touched. Those last ones are candidates to leave the visible column, or to start closed.