What a Two-Ton SUV Doing a Loop Taught Me About A/B Testing

The Stunt That Made Everyone Look

In 2010, Mercedes had Michael Schumacher drive an SLS AMG around the inside of a tunnel—a full 360-degree loop. Flash forward to 2026, and a Chinese EV maker named Voyah decided to one-up that with a two-ton SUV. The Zhuiguang S, which is over 5 meters long and comes with a fridge in the cabin, drove up a circular wall, flipped over the top, and came back down. The video is bonkers. I watched it three times.

But here's the thing that gnaws at me: that video is pure spectacle. It makes you go 'whoa,' but does it make you go 'I'll take one'? Probably not. And that's where A/B testing comes in—not as a buzzword, but as a reality check. While Voyah was busy filming loops, the real question was whether the loop would move cars or just move eyeballs. My guess? It moved eyeballs, and maybe some brand awareness, but sales? That's a different beast.

What A/B Testing Really Means for a Product Launch

Most people think A/B testing is for landing pages and email subject lines. It's not. It's a mindset. You take two versions of something—a headline, a feature, a pricing model—and you let your audience vote with their clicks. The point is to let data, not your gut, make the call.

For Voyah, the A/B test might have been: does a crazy stunt make people remember the car, or does it make them think it's all show and no substance? The company clearly bet on the former. But without a control group, they're just guessing. And guessing is fine if you're playing poker, not if you're launching a $31,000 car.

The Problem with Stunts: No Control Group

In a proper A/B test, you split your audience. Group A sees one thing, Group B sees another. You measure which leads to more clicks, more conversions, more sales. Voyah didn't do that. They just put the video out there and hoped it would go viral. It probably did, but so what? Viral doesn't equal effective. I've seen viral videos that sold absolutely nothing.

The video might have made people talk about the stunt, but did it make them consider buying a family SUV? That's a completely different question. The stunt was memorable, sure. But memorability without a clear call to action is like a headline with no article—it's just noise.

How to Run an A/B Test That Actually Tells You Something

Let's say you're launching a new product. You could do a big, flashy campaign. Or you could run a quiet A/B test. Here's a simple framework that I've used myself:

• Pick your metric. What does success look like? Is it sign-ups, sales, or just brand recall? Choose one primary metric and stick to it. Don't get distracted by vanity metrics like views or likes.
• Create two versions. Version A might be the stunt video. Version B could be a straightforward explainer about the car's features—the 800V platform, the 740 km range, the zero-gravity seats. Make them as different as possible so you learn something.
• Split your audience. Show half your potential customers the stunt, the other half the explainer. Make sure the split is random. Use a tool like Optimizely or even just a simple URL split if you're scrappy.
• Measure and compare. After a set period—say two weeks—look at the numbers. Which version led to more people actually clicking 'reserve' or 'buy'? That's your winner. If the stunt wins, great. If not, you've saved yourself a lot of money on stunts.

Why Most A/B Tests Fail (And How to Avoid It)

One big reason A/B tests fail is that people test the wrong thing. I've seen people spend weeks testing button color when their value proposition is as clear as mud. Or they test 'Fast' versus 'Quick' when the real question is whether anyone understands what the product does. Don't be that person.

Voyah's challenge was that they needed to convince people a big SUV could be sporty. The loop was a bold statement. But maybe the better test would have been to show two different ads: one emphasizing the 475 kW powertrain and the all-aluminum suspension, another focusing on the family-friendly space and the Huawei smart cabin. Then measure which angle resonated more with their target buyers. That would have told them something useful.

Another mistake is not letting the test run long enough. A/B testing requires patience. You need enough traffic to reach statistical significance. If you only show the video to a thousand people, you might get a skewed result. The same goes for product marketing—you can't judge a campaign after a day. I've made that mistake before, and I've learned to wait at least a week, even if it's painful.

The Data Behind the Hype: What the Numbers Say

Voyah's car starts at 223,900 yuan (about $31,000). That's a competitive price for a mid-to-large electric SUV with 800V charging and a 740 km range. But the market is crowded. Rivals like Xiaomi's YU7 and Chery's RX are coming. The real test isn't the loop—it's whether the car offers enough value to stand out. And that's testable.

What if Voyah had run an A/B test on their landing page? One version leads with the loop video, the other leads with a spec sheet. Which one gets more people to click 'reserve'? I'd bet on the spec sheet, but I've been wrong before. That's the point—you don't know until you test.

Applying A/B Testing to Your Own Marketing

You don't need a car that can drive upside down to run a good A/B test. You just need a clear hypothesis and a way to measure. Here are a few practical tips that have saved me from wasting money:

• Start with your highest-impact element. That might be your homepage headline, your pricing page, or your ad creative. Don't start with the footer.
• Change one thing at a time. If you change the headline and the image at once, you won't know what caused the difference. It's tempting to change everything, but resist.
• Use tools like Optimizely, VWO, or even a simple split test on your email list. Google Optimize is being sunset, so don't rely on it.
• Look at the data beyond the first click. Did the winning version lead to more actual purchases, or just more clicks? Clicks are nice, but revenue is nicer.

The Takeaway: Stunts Are Nice, but Tests Are Better

Voyah's loop was a great piece of content. I'll give them that. It got people talking, and it made a point about the car's stability and power. But as a marketing move, it was a gamble. Without A/B testing, they don't know if it actually helped sales or was just a fun video.

In your own work, remember: don't rely on a single big idea. Test, measure, and iterate. The data will tell you what works—and what's just spinning your wheels. And if you ever get the chance to drive a car upside down, take it. Just don't bet your marketing budget on it.