How to Perform a Card Sort


If your store has more than 50 products, you’ll need to categorize them somehow.

Yet categorization is highly arbitrary. How you categorize your set of products will differ from how a fellow team member may categorize them. And those two interpretations may differ substantially from how your customers might think about your products.

You want your navigation to make sense to your customers, and your products to be easily discoverable by them. Time and again, I see conversion rates lift in the segment of customers that simply click to the store page – but you can do one better, by making the store’s landing page as legible and understandable as possible.

The classic way that people do this is through a card sort.

In a card sort, you write (or print) the names of each of your products on a deck of note cards. Then, participants sort them according to either predetermined categories (called a closed sort), or by letting the participants decide categories of their own (called an open sort).

If you want over 250 pages on how to run a card sort, the go-to book on card sorting is called Card Sorting, by Donna Spencer.

Let’s talk about how to run a card sort, and synthesize its findings into your store.

Run a card sort with prospective customers

Ideal card sorts happen with 5-7 prospective customers, in person, in a conference room, over the course of about 2 hours.

Recruitment should happen through an online screener, and you should compensate participants for their time.

If you absolutely must run a card sort remotely, you can use OptimalSort for the purpose – but you will still have to recruit participants.

Open or closed?

If you are very clear about what categories to use in your store, and you’re unwilling to change them significantly, run a closed card sort.

If you’re more open-minded about what products are slotted into what categories, run an open sort.

I tend to recommend open card sorts for online stores, because they potentially allow for more radical reworks of navigation that are likely to have high leverage in any A/B tests you might run. For some stores, though, the answer is more cut-and-dried – especially in industries like apparel & accessories, where people come to expect a certain set of categories.


First, create enough decks of cards as there are participants.

I tend to run index cards through my laser printer, using a custom 3x5 paper template, with the store’s full database of product names run into a mail merge.

Next comes each set of categories. If you have a closed sort, you’ll need to print the categories you expect all of your products to fall into. If you have an open sort, give out stacks of blank cards to write categories on.

Give out a bunch of Sharpies (for the blank cards) and additional blank cards (for other products, ideas, sketches, etc).

Running the card sort

Explain why we’re here, and that there are no wrong answers. The card sort is broken into two phases: the actual sorting, and then follow-up discussion. Discussion is meant to be constructive, and will help participants explain their choices.

Participants should have between 15 minutes and an hour to perform their sort. Generally, assume people are capable of sorting around 5 cards a minute.

Give an extra 15-20 minutes for open sorts, as people will usually have to go through the full list of cards before determining what categories to start with.

Once the sort is done, have people lay their cards out on the table, and start asking questions on 3 things:

  1. Any significant differences between participants’ sorts,
  2. Category choices (for open sorts),
  3. And any choices that might be surprising to you.

Take notes on what the differences are, what the commonalities are, and what that might mean for the final card sort.


Your final output is a card sort, encompassing both categories and products, that incorporates as many participants’ opinions and ideas as possible. The goal is a conensus that removes any outliers and clarifies the group’s overall thinking.

This may be difficult when it comes to an open sort. You may have participants coming up with radically different categories that would be hard to reconcile with one another. Here, understanding the thinking behind each set of categories is critical to knowing where to start when creating your final sort.

Products might need to belong to more than one category – which is hard to pull off on some online store platforms, but essential for findability.

It may also be valuable to create collections of products that are featured elsewhere on the site. Stationery site JetPens does this well, with all manner of collections that speak to very niche needs.

Run a card sort by yourself

Make the cards

First, write each primary navigation item of your existing store on your card, such that you have one per card. Include any links like “my account” or “cart”. Set these aside.

Then, write each primary category of your store on each card, again one per card.

Finally, write each sub-category of your store on each index card, such that you have one sub-category per card.

So you have three stacks of cards. If your store is one of those that puts all of your primary store categories underneath a “store” link, toss the store-only card. (Trust me on this, it’ll make sense in a minute.)

Create the categories

Stores should separate their main navigation into three visually distinct sections: browsing, which gets customers to view collections of products; information, which teaches customers about the brand; and functions, which help customers get through the checkout funnel or work with a previously-submitted order.

On three cards, write “browsing”, “functions”, and “information”. Tape these in a single row on your desk.

The primary sort

Shuffle all of your cards from the first two stacks you made together. Then, go through each one, and determine whether each link applies to browsing, functions, or information. Place each card in a stack underneath its respective category.

The subcategory sort

Finally, take the categories that you placed underneath the browsing stack, and tape each of them in another row beneath everything you’ve done already.

With your final stack of subcategory cards, you’ll repeat step 3 with this row, placing each one underneath the category that makes the most amount of sense for it. Some may apply to multiple categories, which is fine – just go with what makes the most amount of sense for each one right now.

The special case: general categories

Many stores have category pages set up that can be theoretically scoped to the whole product catalog, including:

  • New arrivals
  • Popular
  • On sale
  • Our picks

These should usually be set up as separate primary navigation items, grouped together – and they should also be visually distinct from your regular categories. That way, people won’t need to spend extra brain cells distinguishing special-case categories from the rest of your store.

Designing the results

Now it comes time to sketch out how this will look in the final design.

On desktop and especially mobile, each set of navigation – browsing, functions, and information – needs to be separate and visually distinct.

  • Never combine informational navigation with functional navigation.
  • Never combine either type of navigation with anything focused on showing products.
  • On mobile, show all product categories first in your navigation, and then separate all other types of navigation at the bottom.
  • On desktop, left-align your product categories, and right-align any informational links.
  • Never bury your product categories under a generic link, like “shop” or “store”, on desktop or mobile. Keep them clearly visible and spare customers the added click & confusion.

Testing & Rollout

The goal is for your card sort to shape your overall site navigation. But you still have to test it.

This is best performed with a multi-page test using GET queries, so the new menu’s contents can replace the old one. Hide the new menu by default, and swap the display: styling accordingly.

You’re obviously measuring add-to-cart rate, conversion rate, AOV, and ARPU, but you should also run heat maps on your 5 most-trafficked pages, for both control and variant, to see if there are any significant differences in engagement. Are people browsing more products in a given navigation schema? Are people going back to add more to their product? Are more products being added to cart? Are people availing themselves of more accessory upsells at cart?

These are all things you want to measure, because navigation is a significantly load-bearing element in your store, and overhauling it wholesale can have all sorts of knock-on effects that could cause more harm than good.

← Back to the Blog Check Out Draft’s Store →