The four numbers worth watching, and the ones wasting your time

Open any social media dashboard and you are shown twenty numbers. Perhaps three of them would change a decision. The rest are there because they were easy to fetch.

This is a guide to reading the three that matter, calculating them yourself when a tool will not, and recognising the ways a chart can be perfectly accurate and still leave you with the wrong idea.

1. Reach, and why it is the denominator of everything

How many people saw it. Not how many times it was shown: that is impressions, and the difference matters when one person scrolls past the same post four times.

Reach on its own tells you almost nothing. Its job is to be the denominator: every other number is only interesting as a proportion of the people who had the chance to act.

A post reaching 400 people with 40 likes is doing something a post reaching 40,000 with 400 likes is not, even though the second looks ten times better in a screenshot. The first is talking to people who care. The second was handed to strangers by an algorithm.

2. Engagement rate, calculated the boring way

Engagement rate is likes plus comments plus shares, divided by reach. That is it. If your tool does not show it, the arithmetic takes ten seconds.

What makes it useful is that it is comparable. Follower counts are not: an account of 800 and an account of 80,000 cannot be read on the same axis. A rate can. It lets you compare this month to last, one format to another, and your account to a competitor without pretending you are the same size.

Two warnings. First, calculate it against reach, not followers: the two produce wildly different numbers and most published benchmarks do not say which they used, which makes them useless for comparison. Second, look at it per post before you look at it per month. A monthly average hides the thing you want to find, which is the one post that worked and why.

3. Saves and shares, weighted far above likes

A like costs nothing. A share is someone putting their own name behind your post; a save is someone deciding they will need this later. Both are much rarer and much more predictive of whether something worked.

If you are deciding what to make more of, sort by saves and shares rather than by likes. The ranking is usually different, and the posts that rise are usually the useful ones rather than the pretty ones.

4. Replies you owed and did not send

The one number no dashboard shows, and the only one on this list that measures your behaviour rather than the algorithm’s.

Count, once a week, the comments and messages that asked something and got nothing. Not to feel guilty, but to see the shape of it. Most people discover the backlog is small and concentrated on two or three posts, which makes it a twenty-minute job rather than the looming task it felt like.

Consumer surveys published in 2026 put the expected reply window in hours, not days: Emplifi’s April 2025 study of a thousand consumers found 32% expect an answer within the hour. Whether you meet that or not is a decision. Not knowing how many you missed is not.

The three ways an honest dashboard misleads you

None of these involve anyone lying. They are all consequences of how platform data actually arrives, and knowing them changes how you read every chart you will ever see.

Cumulative counters read as daily activity

Platforms report totals, not deltas. A post has 1,000 views today and 1,150 tomorrow. If software adds those up, it reports 2,150 views for two days, when the real answer is 1,150, of which 150 happened today.

This is the most common quiet error in social reporting, and it inflates by whatever number of times the tool sampled. The tell is a chart where every day looks busier than the last for no reason, or where a single old post appears to be gaining thousands of views a day forever.

Old posts landing on the day tracking started

When a tool starts watching an account, it sees every existing post with its lifetime totals. If it treats those as activity, day one shows a spike of everything the account ever earned, and the following weeks look like a collapse.

If your history begins with an enormous first day, that is what happened. The comparison against “last month” is meaningless until you are a month past it.

Corrections that arrive as losses

Platforms revise their numbers downwards: spam accounts removed, deleted profiles, fraud filtering. A counter genuinely goes backwards sometimes. Software that subtracts naively produces negative activity, and a chart dipping below zero on a day nothing happened is this and not a catastrophe.

What a useful weekly review looks like

Fifteen minutes, once a week, and four questions.

Which post did best by engagement rate, and what was different about it? Not the one with the most likes: the one with the best ratio. Write one sentence about why. After eight weeks you will have eight sentences that are worth more than any benchmark.

Which post did worst, and was it the format or the subject? This is the more useful question of the two, and the one people skip.

Who asked something and got nothing? Clear it. Then look at what they asked: repeated questions are next month’s posts, already written by your audience.

Is the rate trending, ignoring the peaks? One viral post tells you nothing repeatable. The line under the peaks is your actual account.

A note on what your platform will not give you

Some of these numbers are simply not available, and no tool can conjure them. Instagram keeps story statistics for 24 hours and then they are gone. Facebook stopped providing age and gender at Page level in November 2025. Bluesky publishes no reach or impressions at all. YouTube never returns a historical subscriber count, only daily gains and losses.

This matters for a practical reason: it means the history has to be recorded as it happens, by something, or it does not exist. If a number matters to you, start capturing it now rather than assuming you can go back for it later. You cannot.