How to audit your own channel

An 18 August 2026 measurement of one short-video site put its visit-to-subscriber conversion at 0.23 percent. Numbers like that are what an audit is for: not to admire the channel, but to find out which parts of it work and which parts only feel like work. This page walks through a five-step audit you can run on your own channel in an afternoon.

The short answer

An audit is a table, not a mood. List every video with its date, format, and result relative to your channel median, find the one or two formats that carry most of your views, find the videos that broke the pattern in both directions, check the shelf your videos sit on, and write down three changes. Stop at three. An audit that ends with fifteen action items ends with zero, because you will do none of them.

The steps

1. List every video with three columns

Open your video list and build a spreadsheet with three columns: publish date, format, and result. Format means the repeatable shape of the video: talking-head explainer, archive-footage documentary, listicle, screen recording. Result means views relative to your channel median, not raw views, because a video from your first month and a video from last week are not comparable on raw numbers. Divide each video's views by your channel median and record the ratio. Do not skip videos you are embarrassed by. They are data.

The output of this step is one row per video, with a ratio you can sort.

2. Find the format that is carrying the channel

Sort by ratio and look at the top ten rows. In most channels, two formats produce most of the above-median results, and the rest of the catalogue is variation on themes that do not work. Write down those two formats in plain words. Then check the bottom ten rows for the same question in reverse: which formats consistently sit below the median. If a format appears in the bottom ten more than twice, it is not bad luck, it is a pattern.

The output of this step is a sentence: "My channel is carried by X and Y, and Z consistently underperforms."

3. Find the videos that broke the pattern

Now look for exceptions in both directions. A video in a weak format that beat the median tells you something about the topic, the hook, or the packaging. A video in your strong format that flopped tells you the format is not the whole story. For each exception, write one hypothesis about why it broke the pattern. One hypothesis, not five. You are looking for the variable that changed: the title structure, the first three seconds, the subject. Exceptions are the most informative rows in your table because they isolate a variable.

The output of this step is a short list of exceptions, each with one hypothesis attached.

4. Check the shelf, not just the videos

Videos do not perform in isolation. They sit on a shelf made of your title patterns, your thumbnail system, and your descriptions. Look at the titles of your top performers and find the repeated structure: question, number, named entity, tension. Look at your thumbnails as a grid, the way a browsing viewer sees them, and check whether a viewer could tell your videos apart from each other at a glance. Read three descriptions as a stranger would and check whether they say what the video is about in the first line. If you produce Shorts at volume, this step matters more, because the shelf is the only thing a scrolling viewer sees. If you use ViewMade, each render ships with a media credits file naming the source of every clip, which makes the description and citation part of the shelf easier to keep consistent.

The output of this step is a verdict on each shelf element: working, broken, or absent.

5. Write down three things to change and nothing more

Take everything you learned and force it into three changes. Not five, not a list of observations: three changes you will apply to the next videos you make. A change is a sentence you can test: "Open with the outcome in the first five seconds," "Use the question-title structure on every video this month," "Retire the screen-recording format." Attach each change to a check you will run in thirty days. If you cannot say how you will know the change worked, it is not a change, it is a wish.

The output of this step is three sentences, each tied to a date you will re-check.

What this will not fix

An audit explains the past. It does not guarantee the future. The formats that carried your channel last quarter can decay, and the exceptions you found may have been platform-wide trends rather than your own skill. Treat the audit as a snapshot with an expiry date, not a map.

It also will not fix the hardest part, which is that you are auditing your own work. You know why each video was made, what you intended, and what you were going through that week, and all of that knowledge biases the reading. The table helps because a ratio does not care about your intentions. When you feel the urge to explain a bad number instead of recording it, record the number and move on. If you find yourself unable to be neutral, the checklist format helps, because it forces the same questions onto every video whether you like the answer or not.

Where to go next

If you want the five steps above as a fill-in worksheet with the exact columns and questions, the channel audit checklist at /templates/channel-audit-checklist gives you the table ready to copy.

If most of your catalogue is Shorts and you are unsure which analytics numbers matter for the ratio in step 1, the guide at /guides/how-to-read-shorts-analytics explains which metrics to pull and which to ignore.

<!-- faq -->

Frequently asked questions

How many videos do I need before an audit means anything?

Below about twenty videos, patterns are mostly noise: one viral video can drag your median enough to distort every ratio. With twenty or more, the format-level pattern in step 2 starts to be trustworthy. If you have fewer, run the audit anyway but weight your conclusions toward the shelf check in step 4, because title and thumbnail patterns show up sooner than format patterns.

Should I delete videos that underperform?

Usually no. An underperforming video costs you almost nothing once published, and deleting it destroys the data that makes your next audit accurate. The exception is a video that misrepresents the channel, attracts the wrong audience, or violates a platform rule. Otherwise leave it, mark it in your table, and let it teach step 2.

How often should I run this audit?

Quarterly is the working rhythm. Monthly is too frequent, because individual videos need weeks to reach their steady-state view count and you will be reacting to noise. Annually is too slow, because a format that stops working in March will have cost you eight months of output by the time you notice in December.

My channel has one format only. Does step 2 still apply?

Yes, but it moves. If every video is the same format, the carrying variable is not format but topic cluster. Redo step 2 with topic as the column instead of format, and the same sorting logic applies: find the two topic clusters that beat the median and the ones that sit below it.

What if my median is dragged up by one old viral video?

Then use the median of your last thirty videos instead of your all-time median. The point of the ratio is comparability, and a video from two years ago on a different subscriber base is not comparable to anything you publish now. Recompute the ratios with the recent median and the pattern in step 2 will usually get clearer, not blurrier.