Most brands I meet are not short of numbers. They have a dashboard with fourteen of them, refreshed hourly, and no idea which one should change what they do on Monday morning.

That is the real measurement problem. Not too little data — too little consequence. A number you would never act on is not a metric, it is decoration. And decoration is expensive, because it costs you the one thing reporting is supposed to buy: a decision.

So here is the short version of what I look at with the brands I work with, and why. Three signals worth your attention — and three numbers I have learned to leave alone.

A signal is a number that changes your next move

Before the list, the test. A signal has to survive one question: if this number doubled, or halved, would I do something different next week?

Most of what sits on a social dashboard fails that question immediately. You would not change a single decision if impressions rose eleven per cent, and you know it. Keep asking it and the reporting shrinks to something you can hold in your head — which is the point. A measurement habit that needs a spreadsheet to survive will not survive.

Signal one: saves and shares, measured against reach

A like is a reflex. It costs nothing, it happens in the half-second before the thumb moves on, and it correlates with almost nothing you care about.

A save is a decision. Someone has just told you that the thing you made has value later — that it is worth returning to, worth acting on, worth not losing. A share is the same decision made in public, with the person’s own reputation attached. Those are the two behaviours that separate content people consume from content people use.

Measure them against reach, not against followers. Saves per thousand accounts reached tells you how well the idea landed with the people who actually saw it; the same number over your follower count only tells you how many people followed you two years ago.

Ignore benchmarks. There is no universal good save rate, and the ones circulating are usually someone’s client average dressed up as an industry standard. What matters is your own baseline: pull your last thirty posts, work out the rate for each, and find the median. That is the line. Everything above it is a lesson.

Then do the part most brands skip. Look at the five posts furthest above your line and ask what they have in common — not the format, the promise. Nine times out of ten they cluster around two or three ideas. Those are your content pillars, whatever the strategy deck says. The deck is a hypothesis. The save rate is the answer.

Signal two: someone quotes your own sentence back to you

The best measurement tool I have is a folder in my inbox.

Into it goes every message that begins with a version of “I read the thing you wrote about…”. Not “I found you on Instagram” — that is a channel report. I mean the enquiry that quotes a line back at you, or names the specific idea that made the person get in touch.

When someone can repeat your argument before you have said a word to them, the content has already done the selling.

Yes, the numbers are small — three a month, maybe five. Small numbers make people nervous, so they get ignored in favour of impressions, which are large and mean nothing. But this is the only signal on the list that tells you whether the content is changing what people believe, and belief is what shortens a sales conversation from six weeks to one call.

Count them monthly. Write down which piece was quoted. Within a quarter you will have a ranked list of the ideas that move people, and it will not match your engagement chart.

Signal three: people arrive by name

Everything so far happens inside a platform you do not own. This one happens outside it, which is why it is the one I trust most.

Watch the traffic that arrives because someone typed your name: branded search, direct visits, profile visits from accounts that do not follow you. Those people did not stumble into you mid-scroll. They remembered you, left the feed, and came looking. That is the difference between attention you rented and attention you own.

It is also the slowest signal here — it moves on a quarterly rhythm, not a weekly one, which makes it useless for judging a single post and ideal for judging a strategy. If saves are climbing and nobody is searching for you by name, you are producing content that performs but does not accumulate. Usually the fix is not more posts. It is fewer ideas, said more often, in a voice specific enough to be remembered.

Three numbers that lie

  • Follower count. Lagging, gameable, and borrowed. It tells you what happened months ago to an audience you cannot contact directly.
  • A single reach spike. One post going wide almost always finds the wrong people — briefly. Judge reach as a trend across a quarter, or not at all.
  • Average engagement rate. Averages hide exactly what you need to see. Two posts probably carried the month; the mean quietly buries both of them along with the twenty that did nothing.

Twenty minutes, once a month

The review that survives is the one that fits in a single sitting. Mine looks like this:

  • Export the month’s posts with reach, saves and shares. Work out saves per thousand reached.
  • Mark everything above your median. Name the promise each one made.
  • Open the quotes folder. Count the messages, note which piece each one names.
  • Check branded search and direct traffic against the previous quarter, not the previous month.
  • Write one sentence: next month we make more of ___ and stop making ___.

That last line is the whole exercise. If the review does not end in a decision, it was not a review — it was a report, and reports are the most comfortable way to avoid changing anything.

Content that is working rarely announces itself with a spike. It shows up quietly: saves that keep climbing against your own line, messages that quote your own sentences back at you, and a slow rise in the number of people who type your name because they already decided you were worth finding.