A homepage carousel has fans and critics, but neither side has seen your site. Three questions settle it: does anyone click slides after the first, does anyone use the arrows, and do phone visitors scroll past the carousel at all. If the answers are no, the later slides are decoration that costs you space on the first screen. If they are yes, the slider may be earning its place.
This guide covers only the carousel and how to test it on your own data. General rules for what belongs above the fold are in our guide to your website's first screen. Here you will decide what to measure, how to read the result on phone and desktop, and when to replace the slider with a single banner.
What to check before you remove or defend a slider
Carousel debates usually rest on other people's data. Your decision needs four observations from one page, split by device and collected while the slides stayed the same. Each one answers a different question, so none is enough on its own.
Scroll the table sideways to see all columns.
| Question | Data source | Sign the slider is not working |
|---|---|---|
| Does anyone click later slides? | Element ranking on the click map | Links on slide two onward barely appear |
| Does anyone use arrows and dots? | Clicks on carousel controls | Arrows get fewer clicks than the logo or menu |
| Do phone visitors scroll further? | Scroll map for phones | Most views stop at the height of the slider |
| What happens when a slide changes? | Recordings of visits to this page | People scroll on or leave before slide two appears |
Also note what each slide shows and where its link goes, so you can later tell a weak position from a weak offer.
Clicks on each slide link and on the arrows
A click map has one limitation with carousels: the background shows the page as it was captured, with a single visible slide. A colored spot on that image cannot tell you which slide someone clicked. Look at the element ranking instead, where each slide link and each control is listed separately with its click count. How to read the map itself is covered in our guide on how to read heatmaps.
If slide links look alike in your markup, give them readable names. With the data-livechat-track attribute, the Meetlead widget records the click as an event with your name in the visit timeline. Label arrows and dots too.
Slide links: slider-slide-1, slider-slide-2, slider-slide-3. Controls: slider-arrow-next, slider-arrow-prev, slider-dot. Note: the date the slide content changed and the device the data comes from.
Compare slide clicks with other elements on the same page, not with a general benchmark. If the next arrow gets as many clicks as a footer link, visitors are not looking for it. With low traffic, check the sample first: the Meetlead panel marks a map as ready from 50 views of a page on a given device, and our guide on how much traffic a heatmap needs explains how to treat thin data.
Phones: does anyone scroll past the carousel
On a phone, a slider often fills the whole first screen, with no hint of the next section below. Visitors see one photo and must decide whether to scroll or wait. The phone scroll map shows what share of views moved past the carousel. Open it separately for phone and desktop, because the carousel takes up a different share of each screen.
Mark where the slider ends and read the reach just below it. A sharp drop at exactly that point suggests the large image looks like the end of the page. How to tell such a false bottom from a normal exit is explained in our guide to scroll depth. Keep the distinction in mind: scrolling past the slider says nothing about whether anyone saw slide two.
Recordings: is anyone waiting for the next slide
Numbers tell you how many people clicked. A recording shows the first seconds: whether someone watches the slide change, scrolls straight away, or tries to swipe the image and nothing happens. Pick a few phone and desktop recordings of the same page and watch only until the visitor leaves the carousel.
- How many seconds passed before the visitor scrolled past the slider
- Whether the second slide had time to appear
- Whether anyone swiped a slide or clicked the dots
- Whether a click landed on a slide that had just disappeared
- Whether the visitor scrolled back up to find something
The last two points matter most: a click on a slide that just changed leads somewhere unintended, and scrolling back up means information flashed by and vanished. From a point on the click map you can open the recordings linked to that element, if such recordings are retained for the selected range.
In January 2013, Erik Runyon reported carousel clicks on the University of Notre Dame homepage: about 1% of visitors clicked a feature, and 84% of those clicks went to the first slide; that carousel did not rotate on its own. Erik Runyon: Carousel Interaction Stats. It is context from one site more than a decade ago, not a benchmark to measure your own result against.
A simple before-and-after test with a static banner
When later slides are not working, do not remove the carousel blindly. Replace it with a single banner carrying the most-clicked message and compare two periods. A fictional example: a furniture store has four slides, and almost every click goes to the first one, featuring its sofa collection. A banner with that collection and one button is the obvious candidate for a test.
- Record a baseline.Collect slide clicks, scroll reach on phone and desktop, and the number of inquiries from the homepage for the same period.
- Change only the slider.Put a static banner in the same spot. Leave the menu, the offer and campaigns pointing to the homepage unchanged during the test.
- Collect a comparable sample.Count from the day of the change, separately per device, until both versions have a similar number of views and cover the same weekdays.
- Compare three numbers.Clicks on the main button, reach just below the banner and inquiries. Also check recordings for visitors searching for content from removed slides.
A before-and-after test is not an A/B test: traffic and season also shift between periods, so treat the result as a strong hint, not proof. When traffic allows, our guide to A/B testing on your website covers running both versions at the same time. If the change reveals that the button is the problem rather than the carousel, move on to the diagnosis in our guide for when nobody clicks your call-to-action button.
When a carousel is worth keeping
A slider does not have to go if the data shows people use it, for example browsing portfolio projects or photos of a single product. The carousel is then content people browse on purpose, not a way to squeeze several offers into one spot. A carousel with one main message on the first slide is also defensible when the later slides only support it.
- The first slide alone says what you do and who it is for
- Slides do not advance on their own, or they can be paused
- On phones, part of the next section shows below the carousel
- Each slide has one link to a specific destination
- Content from later slides is also available further down the page
WCAG success criterion 2.2.2 (Level A) requires a way to pause, stop or hide content that moves automatically, lasts more than five seconds and is shown alongside other content. W3C: Understanding Pause, Stop, Hide. The criterion is about accessibility, not about whether a carousel brings in more inquiries.
In Meetlead Analytics you open click and scroll maps separately for phone, tablet and desktop, and you can filter recordings of the same page by device. You install the Heatmaps app for a chosen website, and whether recordings are saved and how long their history lasts depends on your plan.
Questions about homepage carousels
Does autoplay help because people see more slides? Not necessarily. A slide shown for a few seconds is not a slide someone has read, and a change mid-sentence breaks the message. If you keep automatic rotation, add a visible pause button and check recordings for clicks landing on slides that are already disappearing.
How long should a static banner test run? Long enough for both versions to have a comparable number of views on each device and to cover full weeks. With low traffic, that can mean several weeks per version. Then rely more on recordings and customer questions than on raw numbers.
