I just finished the biggest data project I have ever run on the Amazon Influencer Program. I pulled more than 11,000 products, roughly 35,000 shoppable videos, and about 13,000 creators, and I ran the whole thing looking for one answer: what actually separates a video that lands in the top carousel slot from a video that nobody ever sees? Not opinions. Not “what worked for me.” Actual patterns across tens of thousands of videos. Some of what I found confirmed things we have all been saying for years. Some of it flatly contradicted advice I have heard repeated in every Facebook group for the last three years. Here is everything that moved the needle, everything that did not, and the exact blueprint I am using on my own uploads going forward.

In This Article

How I Ran This Study (And Why the Adjustment Matters)

Before I give you a single number, you need to understand one thing about how I handled this data, because it is the difference between a study that means something and a study that just tells you what is already popular.

Here is the problem. About 78% of all the videos in my sample were already shot horizontally, and 61% already had a custom thumbnail. So if I just counted raw wins, horizontal would look like it destroys vertical and custom thumbnails would look unbeatable — but only because there are so many more of them. That is not a finding. That is just counting.

So for the tiebreaker factors, I adjusted for how common each attribute already is, and I ran head-to-head comparisons inside the same carousel on the same product. That is why you will see me say things like “57% after adjustment” instead of the raw 82%. The adjusted number is the honest one. It is smaller, and it is the one you can actually trust.

I also tracked two completely separate outcomes, and you need to keep them separate in your head:

  • Ranking — where the video lands in the carousel, measured as average slot position.
  • Engagement — average helpful votes, which is the closest proxy we have to “did a customer actually click this and interact with it.”

Some factors drive one. A few drive both. The ones that drive both are where you should spend your energy.

Video Length Is the Single Clearest Ranking Signal

This was the cleanest result in the entire study, and it was not close. As video length goes up, average carousel position goes up with it. The best-performing length band averaged 2.2 positions in the carousel. That is an absolute knockout compared to everything else I tested for ranking.

I want to be careful here, because I know how this gets repeated. This does not mean go shoot a twelve minute video about a garlic press. What it means is that the very short videos — the sub-60-second stuff a lot of people are pumping out for volume — are consistently landing lower in the carousel. My recommendation from this data is a minimum of 2 minutes, with 3 to 4 minutes as the sweet spot.

If you are one of the people doing 45-second rapid-fire uploads because you are optimizing for count, this is the one finding in the study that should make you stop and rethink. You are trading position for volume, and position is where the money is.

Say the Product Name More Than You Think You Should

This one surprised me with how strong it was. I looked at how many times a creator said the product name and brand name out loud in the video, and the gap between the bottom and the top was significant: videos at the lowest carousel positions averaged 5.9 mentions, while the videos climbing to the top were hitting around 8 mentions per video.

Roughly eight mentions. Say it in the hook, say it when you pick it up, say it when you demo the main feature, say it when you compare it, say it when you wrap up. “This vacuum cleaner.” “The vacuum cleaner.” Say the brand with it when it fits naturally.

The important word there is naturally. Do not stuff it. If you sound like you are keyword-jamming, customers bounce and you lose the engagement side of the equation. But most creators say the product name twice and move on, and the data says that is leaving position on the table.

Horizontal Still Wins, But Not by What You Think

Raw numbers said horizontal won 82% of the time. That number is garbage on its own, because 78% of the sample is already horizontal. After I adjusted for that imbalance, horizontal still wins the number one slot 57% of the time.

57% is a real edge. It is not the blowout the raw number suggested, but well over half is meaningful, and it is strongest at the very top slot specifically. Interestingly, once you get past position one, the vertical videos start showing up closer to 45% — which honestly makes me think Amazon may be doing some deliberate mixing in the lower slots rather than purely ranking on merit.

Bottom line: shoot horizontal. It is a genuine tiebreaker, especially for that top position, and it costs you nothing to do.

Custom Thumbnails Win Twice

This is one of the few factors that showed up on both scorecards — ranking and engagement — which makes it one of the highest-value things on this list.

On ranking: across the 35,000-plus videos and 11,000-plus products, custom thumbnails beat auto-grabbed frames 53% of the time after adjusting for how common custom thumbnails already are. Head-to-head inside the same carousel, custom thumbnails ranked above the frame grab 55% of the time.

On engagement, the gap is much bigger. At the same carousel position, videos with a custom thumbnail pulled roughly 66% more helpful votes on average. In slot one specifically, custom thumbnails averaged around 16 helpful votes. And this holds at every slot, not just the top — regardless of position, the custom thumbnail is pulling more engagement.

Make it bold and high contrast. People click on bold thumbnails. That is not a complicated insight, but 39% of videos still are not doing it, and that is free ground you can take.

The Title Formula the Data Actually Supports

Titles turned out to be almost entirely an engagement play rather than a ranking play, but the effects were consistent enough that I now treat this as a checklist. Here is what the numbers said:

  • Length: 50 to 69 characters. Titles under 30 characters averaged 7.2 helpful votes. Titles in the 50–69 band averaged 11.4. That is a huge spread, and most creators are sitting well under the sweet spot.
  • Include one deliberate ALL CAPS word. With an all-caps emphasis word, average helpful votes came in around 10.3. Without any caps, it dropped to 8.1. Use one — WATCH, NOW, WOW, NEED, BUY. I occasionally do two, rarely more.
  • Frame it as a question or curiosity hook. “Does this actually work?” “Watch this before you buy.” Give people a reason to click.
  • Use punctuation. At least one piece, ideally two. Titles with a colon averaged 16.2 helpful votes, which was one of the strongest single title signals I measured. I usually run a comma plus an ellipsis.
  • Include the word “review.” 9.5 average helpful votes with it versus 8.8 without — about an 8% lift. Small, but free.
  • Skip the word “honest.” This one is a genuine negative: roughly a 25% decrease in helpful votes. Separate analysis last year found the same thing. “Honest review” is dead weight — drop it.

Comparison Videos Are the Biggest Engagement Lever I Found

Nothing else in this entire dataset came close to this number. Comparison videos — the “X versus Y” format — pulled roughly 4.4 times the helpful votes of non-comparison videos in the same sample. In raw terms that is 30.6 average helpful votes versus about 7.

30.6 was the single highest helpful-vote tally of any metric I pulled in the whole study. Nothing was second place by a comfortable margin. It also holds up on ranking, showing at 13.1% in the top positions versus 10% below.

This is exactly why I built the Comparison Video Schedule into Oink. When you have data telling you a format performs four times better and you are still shooting one-off product demos, you are not making a creative choice — you are just not looking at the numbers. The hard part with comparison videos was never the format, it was figuring out which two products to compare and whether that pairing is worth your time. That is a research problem, and research problems are solvable.

Review Videos Own the Top Slots

Review-style videos skewed heavily toward the top of the carousel: 39.6% of slot one versus 29.2% of slot five. That is a clear tilt toward the top, and it is one of the stronger ranking signals in the study.

They also performed on engagement, earning 87% more helpful votes in the same sample — 13.7 versus 7.3.

So genuine review format and comparison format are both strong on both scorecards. If you take one structural change away from this entire study, it is this: stop making generic demos and start making genuine reviews and genuine comparisons.

What Did Not Matter At All

I think this section is as valuable as the rest of the study, because a lot of these are things people obsess over. I tracked all of these and found no meaningful pattern:

  • Chasing helpful votes directly. There is nothing actionable there on its own — it is an output, not an input.
  • Talking speed / words per minute. Nothing. Talk however you talk.
  • Calls to action. No detectable pattern, and honestly there are barely any CTAs in shoppable videos to begin with.
  • Being brand gifted. Zero effect either direction, positive or negative. It showed up plenty in the text and it simply did not correlate with anything.
  • How early you mention the first product. We had enough data to test it and it just did not mean anything.

One partial exception: enthusiasm and energy. In the transcript subsample, position one skewed toward creators who were more positive about the product. But positions two, three, and four were essentially identical to each other. So it is a weak signal, and only at the very top. Be genuinely positive when the product deserves it, but do not fake it — the data does not support forced enthusiasm as a strategy.

The Whole Blueprint in One Place

Here is the checklist. Use it on your next upload.

  1. Shoot at least 2 minutes, ideally 3 to 4. The clearest ranking signal in the study.
  2. Say the product and brand name around 8 times, naturally, spread across the video.
  3. Shoot horizontal. 57% edge at the top slot after adjustment.
  4. Make it a genuine review, or a comparison, whenever it fits. Both are strong on ranking and engagement.
  5. Always upload a custom thumbnail. Never the auto frame grab. Bold, high contrast.
  6. Write a 50–69 character title with one ALL CAPS word, a curiosity hook or question, and at least one piece of punctuation.
  7. Include “review.” Drop “honest.”
  8. Aim it at a product that still has an open carousel slot. This one deserves its own article, and it is coming.

The whole plan in one sentence: a 3-to-4-minute horizontal video that names the product often and compares it to alternatives when you can, published with a bold custom thumbnail and a 50–69 character title carrying an all-caps word, a curiosity hook, and real punctuation — aimed at a product that still has an open carousel slot.

That last point is the one most people are going to skip, and it is the one that decides whether any of the rest of this matters. You can execute this blueprint perfectly on a product where every carousel slot is already locked up by creators with a two-year head start, and you will get nothing for it. Picking the right product first is what makes the other seven items pay.

Every finding in this study points back to the same bottleneck: knowing which products are actually worth your time before you pick up the camera. That is the entire reason I built Oink. Storefront Cross Check shows you what you are missing against other storefronts, Unavailable Video Matching catches the videos quietly losing you money, the Comparison Video Schedule hands you the X-versus-Y pairings that just proved to be 4.4 times better for engagement, and the 5 Pillars system keeps you pointed at products where a top slot is still winnable. Come see the whole toolkit at oinkforinfluencers.com — execute this blueprint on the right products and you will feel the difference within a month.

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