By running a simple AB test, comparing a static carousel with an autoplay carousel, CVC learned that the real gain wasn't in the average. It was hiding where almost no one was looking: from the third banner onward.

CVC is the largest travel operator in Latin America, with over five decades in the business, a presence in more than 1,000 physical stores, and a robust digital operation. At that scale, every interface detail becomes a lever: what looks like a small change on a page, multiplied by daily traffic, turns into revenue lost or won.

That's exactly where this case begins. On one of CVC's highest-traffic landing pages, the top of the page features a banner carousel showcasing the main promotions. In the version that had been running until then, the carousel was static: to browse other offers, users had to click the arrows or the indicators. Functional, but it required effort.

The hypothesis

The product team asked a straightforward question: what if the carousel rotated on its own?

The intuition was clear. Reducing the discovery effort tends to increase information consumption. If the user doesn't have to decide whether to continue seeing more offers, the chances of a promotion catching their attention go up. But intuition isn't data, and CVC treated the question the way it should be treated: as a test.

Every interface decision at the top of the page has a leverage effect. It's worth testing before assuming.

The test setup

The experiment compared two simple variants:

• Variant A (control): static carousel, only advancing when the user clicks.
• Variant B (treatment): autoplay carousel, banners rotating automatically at fixed intervals.

The primary metric was banner engagement, measured by clicks on the displayed promotions. Each banner has two distinct CTAs, one for purchasing on the website and one for purchasing in a physical store, which also made it possible to look at the autoplay effect by conversion flow.

The overall result

Autoplay won, and it won consistently. The auto-rotating variant recorded a 42% uplift in total carousel engagement. The hypothesis held: removing the friction of discovery increases interaction

And the gain wasn't skewed by a single purchase path. When they broke down the result by CTA, both flows pointed in the same direction:

• +42% uplift on the website purchase CTA click rate
• +48% uplift on the in-store purchase CTA click rate

In other words, the effect doesn't favor a specific user profile. Both users who convert online and those who prefer to close with a consultant in-store responded better to the automated format.

Where the test got really interesting

The headline number, +42%, would have been enough to ship autoplay to production and move on. The team, however, took the step that separates a decent test from a well-executed one: they broke down the result banner by banner.

And that's where the story changed entirely.

In other words, the average was hiding three different stories. Autoplay penalizes the first banner, ties on the second, and unlocks a range of offers that, in the static model, simply weren't getting any attention. The 42% aggregate gain, in practice, comes from what happens from the third banner onward.

The overall average confirmed the hypothesis. The position-by-position analysis is what turns the result into a product decision.

Business implications

This finding changes how CVC can operate the promotional showcase at the top of the page.

1. The prime spot may no longer be the first slot

There's an almost automatic convention in e-commerce: the strongest offer goes in the first banner. The data from this test suggests the opposite when the carousel is on autoplay. If the first banner is precisely the one that loses the most from automatic rotation, placing the main promotion there means handing the most valuable piece to the weakest position.

2. "Tail" offers are no longer invisible

In the static format, the third banner forward depended on a user curious enough to click the arrows. With autoplay, those offers began to be displayed organically, and they received over 75% more engagement. This opens the door to using those positions far more strategically, with offer, audience, and message tests.

3. Looking at the average is enough to conclude. It's not enough to learn.

If the team had stopped at "autoplay won," they would have shipped a good change and missed a better decision. The granular reading is what turned the test into the foundation for a digital merchandising strategy.

Next steps: what to test next

Good results open the door to better hypotheses. Off the back of this experiment, three directions are already on the table:

• Initial delay: holding the autoplay for a few seconds before the first rotation, giving the user time to absorb the opening banner. This could recover some of the drop observed in the first position without giving up the gains in the subsequent positions.
• Rotation speed: the interval between transitions is an independent variable. Slower rotations tend to favor reading; faster ones, exposure. There's an optimal point that can only be found by testing.
• Pause on hover: pausing the rotation when the cursor is over the carousel. A clear signal of interest that deserves time. This tends to improve click quality, not just quantity.

Wrapping up

The CVC test is a good example of something that sounds obvious and is rarely practiced: experimentation doesn't end when a winner is declared. It ends when the team understands why the result happened and what it enables next.

Switching the carousel from static to autoplay was, in the end, the easy part. The real value of the experiment lies in learning how users consume a promotional showcase at the top of a high-traffic page and in how many new tests it unlocks.

A 42% increase in engagement is a good headline. Knowing exactly where those 42% come from is what builds the next win.

If you also want to make product decisions backed by data, create your free account and start testing with Croct today.