How to Measure Social Media Performance
Pick a baseline, pick a small number of measures, and know how much data you need before a difference means anything. Most measurement fails at the first step.
Quick guide
Measure against your own history rather than an industry benchmark, choose three measures tied to what you actually want, and treat anything based on fewer than twenty posts as a hint rather than a finding.
- Set your own median as the baseline. Industry averages are not comparable.
- Choose one reach measure, one response measure and one action measure.
- Compare like with like: same format, same channel, same rough period.
- Check the sample size before you conclude anything.
- Review monthly. Weekly numbers are mostly noise.
The full walkthrough is below.
Most social media measurement produces a monthly document that nobody acts on. Not because the numbers are wrong, but because nothing in the document says what to do differently, and nothing says how confident anyone should be.
Three decisions fix that, and the first one does most of the work.
Decision 1: What you are comparing against
The baseline is the whole game and almost everyone gets it wrong by reaching for an industry benchmark.
"The average engagement rate on LinkedIn is 2%" is a number drawn from millions of accounts of every size, in every industry, posting at every frequency. Your account is one of them. The average tells you nothing about whether your Tuesday post did well.
The only baseline that means anything is your own median. A post that did 1.6 times your median did well for you. That is a claim you can act on, and it is available to anyone with twenty posts of history.
Median rather than mean, incidentally, because one viral post drags a mean upwards and then everything afterwards looks like failure.
Decision 2: Which measures
Three, not twenty. One from each layer:
Reach. Did anyone see it. Impressions or views where the network reports them. Worth knowing that many networks do not: LinkedIn, Facebook, Instagram, Bluesky and Mastodon return no reach figure to third party tools at all, so on those channels you are measuring responses without a denominator.
Response. Did anyone do anything. Likes, comments, replies, saves, shares. Weight them: a save or a share is worth far more than a like, because it costs the person something.
Action. Did anyone do the thing you actually wanted. Clicks, sign ups, replies to a call for replies, conversations that started with "I saw your post".
The third one is the only one connected to why you are doing this, and it is the one most reports omit because it is the hardest to attribute.
Decision 3: How much data before it counts
The step everyone skips.
A difference between two posts is not a finding. A difference between two groups of twenty is worth looking at. Somewhere in between is a range where you can convince yourself of anything, and social media data is noisy enough that you usually can.
Rough guide:
| Posts in the group | What you can say |
|---|---|
| Under 10 | Nothing. Genuinely nothing. |
| 10 to 25 | A direction worth testing deliberately. |
| 25 to 100 | A pattern, if the difference is large. |
| Over 100 | A pattern you can plan around. |
Anyone reporting a finding should state which of those they are in. A tool that says "photo posts run 1.6 times your median, based on 24 posts" is more useful than one that just says 1.6, because the second one invites you to reorganise your quarter around noise.
Compare like with like
Three confounders that ruin more analyses than anything else:
Channel. A LinkedIn post and an X post are not comparable in any direction. Different audiences, different mechanics, different reported metrics.
Format. Video and text do not compete on the same scale on most networks, so a mixed comparison mostly measures format.
Time. Two months apart is two different audiences, because your following changed and so did the algorithm.
Hold two of the three constant and you have a real comparison. Hold none and you have a number.
What to actually do monthly
Twenty minutes, four questions:
- Which posts beat my median, and what do they share?
- How many posts is that based on?
- Which measure moved that connects to the thing I want?
- What is the single change for next month?
Write down the answer to four. Next month, check whether it did anything. That loop is the entire discipline, and it beats any dashboard.
FAQs
What is a good engagement rate?
There is no useful general answer. The number that matters is yours from last quarter, and whether this quarter beat it.
Should I track followers?
As a slow background number, yes. As the measure you optimise, no. Followers who never respond are a vanity line and they can grow while everything that matters shrinks.
How often should I look at analytics?
Monthly for decisions. Weekly if you enjoy it, but do not change anything on a week of data.
Why do different tools show different numbers?
Because networks report different things to different kinds of connection, and tools fill gaps differently. Some estimate reach where the network provides none. Prefer the tool that shows fewer numbers and says where they came from.
What if a network reports almost nothing?
Then measure responses and the action you care about, and accept that you have no reach denominator. That is honest, and better than an estimate presented as a fact.
About the author
Founder, Hyvo
Subarna Basnet is the founder of Hyvo, an AI-powered platform helping businesses create, manage, and automate social media marketing. He writes about AI, social media automation, content creation, and marketing technology.