See what is really on your thumbnails, and what to test next.

Add your channel. Base.Tube pulls your latest videos, reads their thumbnails and titles closely and writes them up in plain words: the facts on each, what might be holding it back, and the experiments to run. No score, no click prediction.

Freefree account: email, Google or Discordno YouTube access needed

Sam’s Garage is invented, and its three thumbnails were made for this example. The rest is the app’s real report: the same sections, in the same order, under the same rules. Facts are counted, readings are marked as hypotheses, and nothing is scored.

S

Sam’s Garage

Classic truck restoration · 20 videos analyzed

Public dataExample

Positioning

A one-man restoration channel: old trucks pulled out of barns and brought back to running order, with Sam doing the work on camera. Every video asks the same question: will this one run again?

Each thumbnail uses a different light: warm barn, white workshop, blue night. Two of the three put their title’s words on the image; the third has no text at all.

Evidence board

What is literally on each thumbnail. Facts first; every reading of them is a hypothesis.

videos examined
3videos examined
observations recorded
15observations recorded
experiments proposed
2experiments proposed

Barn Find: Will This 1972 Truck Run Again?

Observed

  • No text on the image
  • Two subjects: Sam in the left third, the truck across the right two-thirds
  • Sam’s face is about 2% of the frame, looking at the camera
  • Warm, golden light with sun rays through the barn boards
  • None of the title’s 8 words appears on the image

Hypothesis, unproven until tested

With no text and a calm, posed frame, the image may read as a portrait of a man and his truck. The question the video asks (will it run?) could be living in the title alone.

Experiment 2: One look across the channel

Experiments to run

In order. Each one is a single change you can prove or disprove.

  1. Experiment 1: Let the image add a second idea

    Hypothesis

    If the words on the image say something the title doesn’t, the pair may tell a viewer two things in one glance instead of one thing twice.

    Variant brief

    Thumbnail: Keep the engine photo but crop in until the engine fills about 40% of the frame. Remove the arrow and most of the workbench. Replace “48 HOURS” with two words the title doesn’t use, such as “SEIZED SOLID”.

    How to measure: YouTube Test & Compare, 2 variants (current and new), run until at least 5,000 impressions per variant.

    Affected videos
  2. Experiment 2: One look across the channel

    Hypothesis

    If every thumbnail shares one kind of light and one text style, the videos may be easier to recognise as one channel when they sit next to each other.

    Variant brief

    Thumbnail: Redo “Will it start?” in the warm light of the barn-find thumbnail. Set the text in the same yellow as “48 HOURS”, at about half its current width, so the truck and its headlights stay visible.

    How to measure: YouTube Test & Compare, 2 variants, at least 5,000 impressions per variant. Then look at the channel page as a whole.

    Affected videos

Swipe file

Packaging worth studying in your niche. Reference material, not a scoreboard.

Similar-sized channelsFound via:barn find truck restorationwill it run after 30 yearsclassic truck engine rebuild
  • Abandoned for 40 Years: Can We Save It?

    Rust Belt Revival

    One subject, the truck, at about half the frame. The title asks the question; the image only shows what is at stake.

  • First Drive After 25 Years

    Two Guys and a Tow Truck

    The same warm light and the same text style on every video, so the channel page reads as one set.

In a real report these are real videos from your niche, with their thumbnails, links and channel sizes. The report then ends with a review queue: which of your recent videos sit far above or far below your median views.

Get this report for my channel

Your reports are saved to your account. Reopening one is free.

Only your packaging: the thumbnails and titles people see before they click. Here is everything in the report, in its order.

Illustration: twenty videos around their median views. The review queue lists those at twice the median or more and at half or less; the free preview reads three closely: the most-viewed, one from the middle and the least-viewed.2× medianmedian½ medianmiddlemost-viewedleast-viewed
read closely in the free previewlisted in the review queueIllustration, not real data.
  1. 01

    Your latest videos, up to 20

    Titles, thumbnails and public view counts, read from your channel. Paste a channel link, an @handle or a channel ID.

  2. 02

    A close read of a few thumbnails

    Your most-viewed and least-viewed videos, plus some from the middle: 3 in the free preview, up to 8 once your own numbers are added. Each is described in countable facts: words of text, how many things compete, how much of the frame a face takes, the light. Those facts are written from the thumbnail and the title alone, without the view counts.

  3. 03

    Positioning

    One paragraph on what your channel is, before any judgment.

  4. 04

    A hypothesis for each video

    A careful guess at what might be holding a thumbnail back, worded as something to test, never as a cause.

  5. 05

    Experiments to run

    Numbered in the order to run them. Each one is a single change, with a brief for the new thumbnail (and the title, when it matters) and how to measure it.

  6. 06

    A swipe file

    Thumbnails worth studying from channels in your niche, with the searches used to find them, labelled by size: similar-sized, mixed, or larger channels (for inspiration, not comparison).

  7. 07

    A review queue

    Which of those videos sit far above your median views (twice or more) or far below (half or less). Descriptive only: views are not adjusted for upload age.

  1. Sign in

    With a free Base.Tube account: email, Google or Discord. You come straight back to the audit.

  2. Paste your channel

    A channel link, an @handle or a channel ID. No access to your YouTube account is needed.

  3. Read your report

    The app shows each step while it works. Your reports are saved to your account, and reopening one is free.

    1. Resolving the channel
    2. Pulling your last videos
    3. Building your swipe file
    4. Observing your thumbnails
    5. Writing up the experiments

Optional, and read-only: Base.Tube reads your YouTube Analytics and changes nothing on your channel. Each video in the audit then gets:

  • Impressions and click-through rateStraight from YouTube. They arrive within 24 to 48 hours of connecting.
  • How long people watchedAverage view duration and the share of the video viewed, right away.
  • Subscribers gained and the main traffic sourceAlso right away.
  • A closer readUp to 8 thumbnails read closely instead of 3.

Rather not connect? Upload a CSV export from YouTube Studio instead. No access to your account; those numbers are labelled self-reported and cover only the dates you confirm.

Every click-through rate says how much evidence is behind it.

The label depends on one thing: how many impressions the video had.

  • not enough dataunder 1,000 impressions. The rate is not shown.
  • directional1,000 to 5,000. Points somewhere, proves nothing.
  • observationalover 5,000. Describes what happened. Still not a cause.

  • A click-through-rate prediction.

    Nobody can predict clicks from an image. A real click-through rate appears only when you connect YouTube, and only from YouTube’s own numbers.

  • A score out of 100.

    No grade and no points. Each video gets facts, a hypothesis and an experiment.

  • A verdict on why a video did well.

    Every reading is a hypothesis. Before a report is saved, its wording is checked for claims that a thumbnail caused a result, and a report that makes one is not published.

  • An A/B test.

    To compare thumbnails on one video, use YouTube’s Test & Compare. The audit tells you what to test and how to measure it.

Something else? Write to support@base.tube.

Your thumbnails and titles written up in plain words, with the experiments to run next. Free, and saved to your account.

Freefree account: email, Google or Discordno YouTube access needed