Your visitors couldn't carousel less

(Reading time: 3m, 17s)

List member Nomi Dayan responded to last week’s look at MoMA’s website (shared with her permission):

Kyle, examining Moma's website is interesting. Some of my thoughts: -- I see why the simplicity is refreshing, but is it possible the website is TOO simple? It doesn't reflect Moma's importance as one of the top museums in the world. Compare to MET website -- Can you describe more about how carousels can be useful?

Lots of interesting topics here — Today, let’s talk about carousels.

I wish I had more positive things to say about website carousels — also called image sliders. The one advantage of a carousel may be that it can help quell internal debate. If you see an automatically rotating carousel on a website, you may be seeing a compromise at work.

Carousels can be a way of deferring tough decisions about what’s most valuable to the audience. Is there a new event coming up? Drop it in the carousel. No need for discussion about how important that event is compared to everything else going on, no risk of ruffled feathers.

What people may not realize is that practically no one sees the information in the carousel.

I’ve pointed you all to my favorite reference on carousels before, which does a good job of demonstrating how bad carousels while also sharing research others have done on the impact and usability of carousels; Here are Nielsen Norman Group’s findings on how auto-rotating carousels annoy users and reduce visibility. Check out Erik Runyon’s tests on Notre Dame’s website:

  • Homepage visits: 3,755,297
  • Percentage that clicked a feature: 1.07%

Percentage of total clicks for each position:

  • Position 1: 89.1%
  • Position 2: 3.1%
  • Position 3: 2.4%
  • Position 4: 2.8%
  • Position 5: 2.6% feature click-through rates, via  Erik Runyon feature click-through rates, via Erik Runyon


So, 0.001% of all visitors clicked on anything beyond the first slide in the carousel.

Another reason you might add a carousel is to satisfy someone who believes that an automatically rotating carousel makes the homepage more dynamic or visually appealing. Again, this is all about satisfying decision-makers assumptions about the audience’s expectations. Employees sometimes assume people visiting a website are evaluating it in the same way they are. And they want their organization to have a website they can be proud of — and that’s understandable. They want a design that both stands out (“dynamic”) and that conforms to what others are doing. A carousel seems like the perfect solution.

What we forget is that people come to websites to complete tasks — they want to learn about the museum, see when it’s open, evaluate membership options, or buy tickets. They do not come to evaluate how “dynamic” the website looks.

No one ever said, “I couldn’t find what I was looking for on the website, but wow — the homepage is really lively.”

So, if you, dear reader, have a carousel on your website, I won't suggest you remove it today based on best practices. Instead, see for yourself.

Here’s what you might try:

  1. Install something like Hotjar (free with some limitations) on your website and run a heat map on the page with the carousel. This will show you how many people are clicking on your carousel to stop it or view previous/next slides.
  2. Wait for the results.
  3. Download the resulting maps
  4. Now, turn off the auto-rotation function on the carousel
  5. Run another map
  6. Wait
  7. Download the results

Now, compare the maps.

Do people seem to interact more or less with the carousel when the rotating function is off?

It’s a tricky question because you can’t gauge intent or emotion through a heat map — are people clicking with interest or out of frustration?

The more important question is whether people are clicking much at all in either scenario. Be prepared for the possibility that a tiny percentage of people are engaging with the carousel at all.

Before you start the test, set a threshold. For example, “If the carousel feature sees less than 2% of clicks on average across both maps, then we’ll have determined that a carousel is inconsistent with the way our visitors want to learn about our offerings, and we’ll try another way.”

Write it down and share it with your colleagues before you begin.

I should note that this is a quick approach. Ideally, you'd use mixed methods to be more confident in the results — perhaps layering in user testing, for example. But if you haven't done much to measure user behavior up till now, a simple exercise like this can be a step toward making more informed decisions. And, in the case of carousels, you may not need a huge sample and triangulation to see that carousels not a useful interaction model for your visitors.

If any of you do try this, email me and let me know what you find. I’d love to hear from you.

Till tomorrow,