# Your best post probably wasn't

Dean Fankhauser, Founder · Published 8 Sept 2026 · Learning

_The number you screenshotted at hour six is not the number at day seven, and half your platforms never report the metric you ranked on. How to read a week without fooling yourself._

## The short answer

- A number seen at hour six is a fact about hour six, not a fact about the post.
- TikTok never reports watch time or retention, and YouTube Shorts retention arrives two to three days late.
- Treating a missing metric as zero quietly ruins every comparison you make after it.
- Change one thing per account per week, or the result tells you nothing you can use.

Someone screenshots the number at hour six and drops it in the channel. Everyone says nice. By Sunday another post has quietly gone past it, and nobody checks, because the celebration already happened and celebrations do not get revised.

That is the trouble with "our best post". It is a claim about a week made from a glance at a Tuesday, and it is the post people then imitate for a month.

## The screenshot problem

Views keep moving. A short video is not finished at hour six, it is barely started, and the number you saw was true then and is a different number now. Most content strategy gets written in the gap between those two facts.

We store snapshots rather than one figure. Every measurement keeps its source and the time it was observed, so a post hands you all of its snapshots instead of the newest one standing in for the story. That sounds pedantic until two posts swap places between Wednesday and the following Monday.

The fix is dull. Decide when you look, and look then. Hour six only ever tells you about hour six.

## Every platform shows you a different slice

This part is not philosophy, it is plumbing.

TikTok never exposes watch time. Not late, not partial. The same goes for average watch and completion. What you get is views, likes, comments, shares, and a saves figure our recorded evidence has not verified.

YouTube Shorts does give you watch time, and its retention fields land two to three days after the post. A Monday video has no retention on Monday or Tuesday, so ranking on Wednesday morning puts it against a Thursday post that has not caught up. YouTube never exposes saves.

Instagram Reels reports watch time. Average watch and completion there are estimable, meaning suitable evidence could support a derived value rather than the platform handing you one. Saves are unverified there too.

Follower delta is modeled as estimable on all three and currently comes back null everywhere, so null is what we return. Other destinations Postdom can publish to have no metric classified against them yet, and every field there returns null with a stated reason. An empty row is not a zero row.

Add that up and "our best post across all platforms, by retention" is not a comparison. It is three questions with one answer stapled across them.

## A blank is not a zero

The most expensive line in any analytics pipeline is the one that quietly turns a missing value into `0`.

It looks like tidiness. It is a lie with a spreadsheet's confidence. TikTok has no watch time and will not have any next month, so every TikTok post scores zero forever, your average sags every time you post there, and your best-post list becomes a list of posts that were not on TikTok. Nothing errors. You just believe something false and build a schedule on it.

So we hold a line that costs us a prettier grid. Zero is a measurement. Null is a boundary.

Zero means the platform supplied a number and it was zero. Somebody published, nobody watched, and that is real evidence, so it stays. Null means no evidence-backed number arrived, and the reason travels with it. Never and unverified always carry a reason. Some other nulls do not, and we would rather say so than paper over it.

You lose the comfort of a full table. It was a false comfort.

## How to read one week without fooling yourself

Three rules, all of them about narrowing.

One account. Comparing posts across accounts mostly measures the accounts: audience size, posting history, whatever the platform decided that week.

One window, preferably a closed one. You can read a rolling 7 or 30 days on an account, the latest two completed account-local weeks, every snapshot of one post, or the outcome of an approved plan. Completed weeks are the ones worth arguing about; a week in progress changes shape while you talk.

One metric that exists on every post in the set. This is the rule people skip. If half your week went to TikTok, retention cannot be your weekly metric, so views is. That feels like a demotion. A weak metric with no holes beats a strong metric missing from half the rows.

_Three questions for a week of posts_

| Question | Ask it of | What it changes |
| --- | --- | --- |
| Did it publish? | Every post in the window, with its outcome and failure reason | Separates a content result from a delivery problem |
| Is the metric real for this platform? | Each destination, against its availability state | Decides which posts can be ranked at all this week |
| What moved since last time? | The same account, over the last two completed weeks | Turns a number into a direction |

Fifteen minutes on a Monday, in that order. The first question catches more than you expect.

## Ranking honestly

Postdom ranks one account by one evidence-backed metric. A post without that metric does not get a zero and quietly lose. It moves to an excluded list with the reason attached.

Say two cuts of the same launch video went out, one with a measured watch time and one whose latest snapshot has no evidence-backed watch time at all. The first ranks. The second sits in the excluded list with that reason next to it. It might be the better video. The ranking says so rather than sorting it to the bottom.

The excluded list is the useful half of the report. A long one is a message: you asked a question your platforms cannot answer this week. Change the metric, not the data.

Null is never used as zero to manufacture a winner. A manufactured winner is worse than no winner, because people act on it.

## Change one thing

Here is the other way a good week gets wasted. You decide the hook is slow, so you rewrite it, cut eight seconds off the video, and move publishing from morning to evening. The numbers go up. Which change did it?

You have a result and no knowledge, and the whole bundle gets filed under "what works" and repeated for a quarter.

One change per account per week, everything else held still. For example: keep the length and the publish time exactly as they were, change only the first two seconds, then read views on that account over the completed week.

## What a good next-week note looks like

The note at the end of a week is not a verdict. It is a description of what happened, small enough that whoever plans next reads the whole thing.

Postdom's digest is built that way: what ran, what published or failed and why, top posts by the metrics that were available, week-over-week movement, and where the coverage gaps were. It does not tell you what to post next. That is a gate we put in on purpose. Recommendations off thin evidence are confident noise, and once a digest starts guessing you stop trusting the parts that were measured.

Your own note can be this small:

_A week note small enough for whoever plans next_

```json
{
  "keep": ["same publish time", "one account per experiment"],
  "change": "first two seconds of the hook",
  "why": "views rose on the two posts that opened on a face; watch time is unavailable on TikTok, so this is views only"
}
```

The "why" carries its own limits, and that is what protects you later. Write "hooks with faces work" instead and you will believe it in three weeks, long after you have forgotten it came from two posts and one metric.

So was your best post really your best? Maybe. If it was measured at hour six, ranked against platforms that report different fields, and compared with missing values filled in as zero, you do not know. You have a screenshot and a feeling. Read the week properly and it tells you something smaller and truer.

> **See what actually came back**
>
> Run one week through Postdom and read the digest with the availability column open.
>
> [Start free](https://app.postdom.com/signup)

## FAQ

**When should I look at a post's numbers?**

Later than you want to, and at the same point every time. Pick a window, let it close, and read it against last week's same window on the same account. Hour six is fine for catching a publish failure, not for judging a post.

**What if most of my posts are on TikTok and I care about retention?**

Then you cannot have it. TikTok never exposes watch time, average watch, or completion, and no tool can retrieve a field a platform does not give out. Measure views, shares, and comments there instead, or run retention experiments on Reels and Shorts, allowing for the delay on Shorts and a different audience answering.

**Does Postdom tell me what to post next?**

No. It returns descriptive evidence with the observation context, availability state, and exclusions attached, and your agent decides what to do with it. Prescriptive recommendations are gated on purpose until the evidence behind them deserves the word.

---

Canonical: https://postdom.com/blog/your-best-post-probably-wasnt