The Monthly Review That Actually Changes What You Make

A week of short-form data is mostly noise. One video gets a push and the week looks excellent; the next gets nothing and the week looks like a failure. Reviewing weekly therefore turns into mood management, and mood management produces decisions like "post more" and "make better hooks", which change nothing.

A month is long enough for the noise to average out and short enough that the lessons are still useful. This is a routine that fits in about forty-five minutes once a month.

Record before you look

Before opening any analytics screen, write down what you published last month: how many videos, which formats, which topics, which lengths, what hooks you used, and roughly how long each one took.

Doing this first matters because the alternative is explaining the numbers with whatever you remember, and memory is not neutral. It favours the videos you enjoyed making and quietly deletes the ones you rushed. The record is the only honest witness.

The five questions

Which three videos held attention best, as a percentage, among videos of similar length? Percentage, not views, and matched by length, for the reasons in which metrics matter.

Which produced the most shares and saves per thousand views, and what did they have in common? Shares are the only engagement that costs the viewer something, so this question usually finds the videos that were genuinely useful or genuinely recognisable.

Which topics did I keep making that never held attention? Look for anything you published twice or more that never cleared your median. That is a topic you find interesting and your audience does not.

What did I spend the most time on, and did it show? Sometimes the answer is no, and that is one of the most valuable findings available. Effort spent where nobody can perceive it is effort you can reclaim.

What did I avoid making because it felt risky? Usually one or two ideas per month get shelved for being unlike the rest. They are also the only place a new format can come from.

Deciding what to stop

Retirement needs a rule, or you will keep making things out of habit. A workable one: retire a format after two months of data where it never clears the median for its length, regardless of how much you like it.

Two months, not two videos. A single weak video is noise. Two months of consistently below-median output is a verdict.

Retiring is not deleting. Take the format off the schedule and keep the files; formats come back when the audience changes.

Deciding what to repeat

Repeat the shape of the top performers, not the exact subject. If the winners share a structure (a question, a single claim, a demonstration), make two videos next month in that structure on different subjects.

Two, not ten. A format that gets overused looks tired within weeks, and the audience that liked it first is the one that notices.

Choosing one experiment

Every month gets exactly one experiment, stated in one sentence before the month starts, with the measure and the threshold included:

Next month, three videos open with a number instead of a question. If two of the three hold more than the month's median at three seconds, numbers become a default hook.

Written in advance, that sentence cannot be renegotiated afterwards. Unwritten experiments always turn out to have worked, because the goalposts move while you are looking at the graph.

The trap of chasing last month's winner

A format can be exhausted without failing. If you have made the same kind of video eight times and the numbers are gently declining, the format is not broken; the audience has simply seen it. The fix is variation inside the format, not more of the same, and certainly not a retreat to the last thing that worked two months ago.

Keep the record honest

The whole routine rests on one file with one line per month: posts published, median percentage viewed, shares per thousand, follows per thousand, the experiment, and whether it was confirmed.

Keep the failures in it. A record that only contains wins is a record that will convince you, twelve months from now, that you have always known what you were doing. The value of the file is that it survives your optimism.

Comparing months with different volumes

Months rarely have the same output. Ten videos one month and three the next will move every total, so compare ratios and the median rather than the sum, and say which number you are comparing before you start.

The exception worth tracking as a total is search traffic, because it accumulates: a month with fewer posts can still beat the previous one on search, and that is a genuine improvement rather than a volume artefact.

Seasonality also makes adjacent months incomparable. If your audience is tied to a season, a holiday or a school term, compare against the same month last year as well as against last month, and keep that comparison in the record so you are not surprised by it annually.

The numbers to write down, and the ones to leave out

A monthly line should be short enough that you will still fill it in during a busy month: views, median percentage viewed for the month's videos of similar length, shares per thousand, follows per thousand, and the count published.

Leave out anything you cannot act on — total impressions, profile views without a follow conversion beside them, watch time across mixed lengths, and any platform-specific score whose definition you would struggle to explain.

The test from the metrics guide applies here too. If doubling the number would not change next month's plan, it does not belong on the line.

A month when nothing worked

Some months produce nothing. Every video underperformed, the experiment failed, and the honest line in the record says so.

That is still a result, and it belongs in the same format. A record of nine good months and one bad one is more useful than a record that skips the bad one, because the bad month is the only evidence of what your floor looks like. It also stops the next bad month from feeling like the end of something.

Two questions are worth asking in a month like that, and neither is "what should I post more of". Did something change outside my control — a platform update, a seasonal dip, a week off? And did I actually execute the plan, or did the month go to reactive posting? Those have different answers, and only the second one is fixed by working harder.

The quarterly pass

A month cannot see slow decay or slow compounding. Once a quarter, read the monthly lines as a series and ask a different set of questions: is median percentage viewed drifting down, is search traffic accumulating, is follow conversion improving, and how many formats were retired versus introduced.

That is where the shape of the account becomes visible. Monthly reviews choose what to make. The quarterly pass decides whether the account is going where you meant it to.

If the month produced one retired format, one repeated structure and one honest experiment, the review worked, whatever the view count did.