A 55% drop that was not real: how partial months lie in monthly reporting
Short answer
If a social export stops mid-month, the final month's totals are automatically lower than complete months, which reads as a collapse. Trend on median impressions per post instead.
Last week I ran a 384-post TikTok export through my reporting tool and it told me the account was down 55% month on month. That number was rubbish, and the reason is worth knowing because almost every monthly social report has it.
Where the fake drop comes from
The export ended on the 24th. July was three quarters of a month being compared against a stack of complete ones. Of course the total was down — there were fewer days in it.
This is not an exotic edge case. It is what happens every single time someone pulls a report before the month closes, which is most of the time, because the report is usually wanted *during* the month.
The fix is to stop trending totals
A monthly total answers "how much did we do", which bundles two very different things: how often you posted, and how well each post did. Separate them and the artefact disappears.
Median impressions per post is immune to all of it. It does not care how many days the month had, how much of it your export covers, or whether you posted nine times or ninety.
Run the same data that way and the picture changes:
| Measure | What it said | Is it real? |
|---|---|---|
| Total impressions, month on month | Down 55% | No — the month is 77% complete |
| Median impressions per post | Down 30% | Yes |
Down 30% per post is a genuine finding. Something changed in how the content is landing. Down 55% would have sent someone hunting for a catastrophe that never happened.
Why the median and not the average
Social distributions are savage. On that same account the top 38 posts carried 53% of all impressions. One breakout post drags the mean up enough to hide a declining floor entirely.
The median tells you what a typical post does, which is the thing you can actually plan against. The mean tells you how lucky you got.
Use both, but never trend on the mean alone.
What to do on Monday
- Trend on median per post. Keep totals for the "how much did we ship" question, and label them as such.
- Mark the partial month in the table. A footnote costs nothing and stops the misreading at source.
- Report cadence separately from performance. If reach fell because you posted half as much, that is a resourcing conversation, not a content one — and the two get fixed by completely different people.
Most reporting arguments I have sat in came down to two people looking at the same number and meaning different things by it. Splitting volume from performance ends most of them.
Key facts
- An export pulled mid-month produces a final month with fewer days, so its totals are not comparable with earlier months.
- A 384-post TikTok export showed a 55% month-on-month drop in total impressions that was entirely an artefact of the export ending on the 24th.
- Measured as median impressions per post, the same data showed a 30% decline — a real but much smaller performance signal.
- Median per post is unaffected by month length, month completeness, or how many times you posted.
Common questions
- Why does my social reporting show a big drop in the current month?
- Most likely because the month is incomplete. If your export was pulled before the month ended, its totals cover fewer days than the months you are comparing against, which produces a decline that is not real.
- What should I trend instead of monthly totals?
- Median impressions per post. It is unaffected by month length, how complete the month is, or how many times you posted, so it isolates performance from volume.
- Should I use the mean or median for social performance?
- Median. Social reach distributions are heavily skewed by outliers — on one account analysed, the top 10% of posts carried 53% of all impressions, which drags the mean far above what a typical post achieves.
Want this run on your own numbers?
The audit tool applies all of it to your export, free, in your browser.