# Test the thumbnail before you re-cut the video

If a video underperforms, test the thumbnail first: play rate is plays divided by impressions, and the poster frame decides it before anyone has seen a second of your edit. Two variants, one meaningful difference, run until each has a few hundred impressions, keep the winner.

- Section: Blog (https://videokr.com/blog)
- Page: https://videokr.com/blog/thumbnail-testing
- Last updated: 2026-08-18

## Why the thumbnail first

Re-cutting a video is a day. Swapping a thumbnail is a minute. And the two metrics are cleanly separated:

- **Play rate** — a thumbnail, placement and headline problem.
- **Retention** — an edit problem.

If nobody presses play, nothing about your edit matters yet.

## What to test

One meaningful difference per test, not a redesign:

- **Product doing the thing** versus a person's face.
- **Text on the thumbnail** versus none — and if text, three or four words, legible at 320px.
- **The end state** (the finished result) versus the starting state.
- **A high-contrast frame** versus a dark or busy one.

## How to run it

1. Two variants, split by impressions on the same placement.
2. Wait until each variant has a few hundred impressions. Ten plays is not a result.
3. Compare play rate, not plays — traffic is never split perfectly evenly.
4. Keep the winner, and note *why* you think it won; the pattern usually transfers to your next video.

Videokr supports A/B thumbnails per video and reports play rate per variant alongside the rest of the [analytics](/docs/analytics).

## Frames that reliably lose

- A random mid-sentence frame of a talking head, mouth half open.
- A slide of bullet points.
- A logo.
- Anything unreadable as a small card in a dark-mode page.

## When the thumbnail is not the problem

If play rate is healthy and retention falls off a cliff in the first ten percent, the thumbnail is writing a cheque the video does not honour. That is the one case where a great thumbnail is a warning rather than a win — and the fix is the opening of the video, not the image.

## FAQ

### How many impressions do I need to call a thumbnail test?

Enough that a few percent difference is not noise — a few hundred impressions per variant is a practical floor for a marketing page.

## Related

- [How to read video analytics and act on them](https://videokr.com/guides/video-analytics)
- [How to build a video landing page that converts](https://videokr.com/guides/video-landing-page)
- [Player, skins and branding](https://videokr.com/docs/player)
- [A video SEO checklist you can finish this afternoon](https://videokr.com/blog/video-seo-checklist)

---
Videokr — Videokr is hosted video for marketing sites: upload or link a video, brand the player, capture emails inside it, embed it anywhere and read second-by-second retention. Free tier forever, $69 lifetime, or metered plans.
Plans: https://videokr.com/#pricing · Full reference: https://videokr.com/llms-full.txt