About two months ago the upper carousels on Amazon product pages chose chaos. You could look at a product, refresh the page, and get a totally different set of influencer videos. Refresh again, different videos again. We all knew Amazon doesn’t do things for funsies, so something was being tested. Then, about two and a half weeks ago, I noticed the opposite: refresh, same videos. Refresh again, same videos. Something had changed, so I did what I did a few months back and ran a full carousel analysis. This time I checked almost 900 products, over 113,000 individual carousel slots, and the results are not what I expected.

In This Article

How I Ran This Analysis

Credit where it’s due: McKay (go follow him, he puts out great stuff) noticed the same consistency shift and started testing a handful of products over an extended period. He found the top two slots had a stickiness that slots three, four, and five didn’t. That pushed me to do a proper follow-up to my earlier analysis, where longer videos and custom thumbnails were getting a clear edge.

Here’s the methodology. I checked a couple thousand products and threw out anything without an upper carousel. That left 893 products. Of those, 412 had fewer than five influencer slots filled, which by itself tells you there’s still a lot of open real estate out there. The remaining 481 had a full five-slot carousel. Out of that 481, 301 would rotate videos at some point when I refreshed, and 180 never rotated no matter how many times I hit the button. Of the 301 that rotated, 224 rotated through all 100 of my refresh attempts.

In total I graded about 113,000 slots across almost 4,000 lineup snapshots and 1,500 logged visits. What I was measuring is fair share: how often a type of video shows up in a carousel and which slot it takes compared to what you’d expect from pure volume. That weighting matters, because most people tag only one product and more people now use custom thumbnails than don’t. If I didn’t weight for that, those categories would look dominant just because there are more of them.

MABS2: What I Think Just Happened to the Carousels

Every single influencer carousel I looked at now carries a label called MABS2. My hypothesis is that it stands for multi-armed bandit system. It’s a very common testing framework Amazon uses in a bunch of different areas, and the concept is simple: picture a gambler at a row of slot machines. Try a little of each, then keep playing the ones that pay off. The payoff here is clicks, watch time, or sales.

That explains the two months of chaos. Amazon was throwing everything against the wall, serving different lineups to every customer and measuring what moved the needle. Now that the testing phase appears to be over, the consistent results we’re seeing feel intentional. They gathered an enormous amount of data during Q3 and the current lineups are the output. I could be wrong about what the acronym means, but it fits what we all watched happen.

Do Longer Videos Still Win?

Not like they used to. A few months ago longer videos had a real edge. Now, when I look at visibility (how often a video appears at all), longer videos do still improve as length increases, but the gap has collapsed. Videos over two minutes win about 53% of the time. Videos under 30 seconds are at roughly 49%. That is almost a coin flip. And here’s the kicker: under-30-second videos actually outperformed the 30-second-to-one-minute bracket and tied the one-to-two-minute bracket.

On position, there’s a slight lean toward longer videos in slot one, but the biggest gap in the whole length breakdown was under 4 percentage points. Outside of slot one it’s essentially nothing. In slot two, videos two minutes and up actually did slightly worse than under-30-second videos. Only in slot five did the longer videos clearly work their way down. So is video length still the key? No. If anything, shorter videos have an opportunity they didn’t have before.

Custom Thumbnails Are Now a Coin Flip

Same story. Once I weighted for the fact that more people are using custom thumbnails now, the fair-share edge for a custom thumbnail came out at about 51%. For position, there’s a barely-over-one-point edge in slot one, a tie in slots two through four, and in slot five custom thumbnails actually underperformed. There is no ranking advantage for having a custom thumbnail right now.

I’m still doing them anyway, and I’ll explain why below. But if you were making custom thumbnails because you thought they got you into the carousel more often, the data doesn’t support that anymore.

More Videos Means More Visibility but a Worse Position

One of the more common strategies lately has been to stack a product with a bunch of videos. Does it work? Partly. Every additional video gives you more visibility on that product, and if you have more than five videos on a product you’re about 67% likely to hold a slot in the carousel at any given time.

But it comes with a cost, and this was one of the largest gaps in my entire analysis. Creators with more than five videos on a product get slot one far less often. If you make one video for a product you have the highest chance of landing in slot one, around 21%. Five or more videos drops that to about 17.5%. For slots three, four, and five, the multi-video creators had a real edge. So you buy exposure, but you give up the top spot.

This is exactly why I built the Comparison Video Schedule into Oink. Comparison and stack videos let you cover more products without spamming five near-identical videos onto one listing, which Amazon’s own documentation tells you not to do anyway.

Tagged Products and Video Age

How many products you tag: no edge whatsoever. The vast majority of people tag a single product. There was a tiny blip for videos tagging four or more products, but the sample was 172 videos versus roughly 3,500, so I have no real takeaway there.

Video age was more interesting. Only 854 of the 5,954 videos I checked, about 14%, even displayed an upload date. Among those, videos over two years old had the strongest overall share at 1.29 (where 1.0 is fair share), and the runner-up was videos under 30 days old. A video over six months old beat a newer video on the same product 64% of the time. There’s nothing you can do about that; you can’t make a video older. But it does mean your back catalog is still working for you.

The Small Pool at the Top

McKay flagged this too, and I saw it everywhere. When a carousel rotated, the top two slots kept cycling through the same very small pool of influencers. One person would bounce out and right back in, and on a lot of products someone was almost always sitting in slot one or two. Meanwhile the lower half of the lineup was a free-for-all, with some creators getting 80% visibility and others getting single digits.

So this isn’t just about which videos you can see in the carousel. Amazon is clearly prioritizing some videos and deprioritizing others before they even get a shot. That lines up with the earnings drops people started reporting three or four weeks ago; videos that used to get decent visibility got pushed toward the bottom.

What I Am Changing in My Strategy

If you’re still locked into the idea that longer videos are the biggest needle mover, I’m personally letting that go. I’m not abandoning longer videos; I’m adding shorter ones alongside them. Here’s what I’m locked in on going forward:

  • One good anchor video per product, the thorough one that does the heavy lifting.
  • A 30-to-45-second vertical short on top of the anchor. Not a bad short video, a good one that serves the buyer and also works on other platforms.
  • Stack videos and comparison videos to cover more products without going five-deep on any one listing.
  • Staying under five videos per product so I don’t sacrifice my shot at slot one.
  • Custom thumbnails, still. Not for ranking, but because when I’m sitting next to four other influencers in a carousel I want to be the one who gets the click.
  • Tagging one product per video, same as always.

The question we haven’t asked ourselves yet is what a good short Amazon influencer video actually looks like. We know what a good long one looks like. Under 30 seconds, what causes someone to buy? That’s the next conversation.

Quick Summary of the Findings

  • Longer videos show up more: barely. Over two minutes wins 53% of the time. Better position: only a slight slot-one lean.
  • Shorter videos show up more: no, but they’re up to 49% and have closed the gap almost entirely.
  • Custom thumbnails: 51% visibility, flat across all five positions. A coin flip.
  • Multiple videos per product: yes for visibility, but a clear drop in slot-one placement.
  • Number of tagged products: no effect.
  • Video age: older videos still hold strong share. Nothing you can control.

I don’t expect this to last forever. Amazon invested heavily in this test and we’re seeing the output heading into Q4. Go eyeball your own carousels and you’ll see it: 12-second videos sitting in slot one over and over. The main thing hasn’t changed, though. Keep creating good content on products that are actually converting. If you want help finding those products and planning your anchor, short, stack, and comparison videos, that’s exactly what Oink is built for.

Ready to build a smarter video lineup for Q4? Grab the Oink extension at oinkforinfluencers.com and start finding the products worth making content for.

I'm an Amazon Influencer since 2023 and blogger since 2021. My goal is to help you succeed in the Amazon Influencer Program.

Similar Posts