Analytics that change the next post.

A metric answers a question whether or not you asked one. Reach answers “how many people saw this?” Saves answer “did anyone want to come back to it?” Link clicks answer “did anyone take the next step?” Reported together with no question, they answer nothing in particular.

So write the question down first, in the words the business would use. For a restaurant running a campaign for Sunday lunch, the question is not “how was engagement?” It is closer to “are these posts getting people to look at the booking page?”

The second version can be wrong, argued with and improved. That is what makes it useful.

The owner, the person writing the posts and the agency reporting on a campaign all look at the same account. They need different things from it, and a single dashboard for all three serves none of them well.

• The owner.Wants to know whether this is bringing people in. Reach, link clicks and the posts that did best are usually enough.
• The writer.Wants to know what to repeat. Saves, the channel and the day matter here, down to the single post.
• The campaign.Has its own question: did this idea land? That needs results for those weeks and those posts only.

This is what custom views in Kanso Analytics are for: a saved choice of tiles for each reader, opened again next month with fresh numbers. The principle holds in any tool, including a spreadsheet. Decide who is reading before deciding what to show.

Suppose two posts opened with a question and were saved about twice as often as the other four. That is worth noticing. It is also two posts.

A pattern in six posts is a reason to try something, not a rule to follow.

Small accounts rarely have enough posts for certainty, and they do not need it. Write the pattern down as a guess, with the posts it came from. Then plan one or two posts that test it and look again.

The same care applies to suggestions from AI. A tool can spot that question openings were saved more often. It cannot know that one of those posts also went out on the day of a local event. A suggestion is useful when it shows the posts behind it and says when the sample is small, so a person can judge it.

A report for a client or an owner should be readable in two minutes: what the question was, what happened, what you will do next. One sentence each is often enough. Charts support those sentences; they do not replace them.

Then carry the decision somewhere it will be used. “Open two posts with a customer question” belongs in next week’s plan. “Thursday evening seemed to work” belongs next to the schedule. A lesson that stays in the report is forgotten by the next one.

• The question.What were these posts supposed to do?
• The result.The one or two numbers that answer it, compared with last period.
• The guess.What seemed to work, and how many posts that rests on.
• The next step.What changes in the next plan because of it.

If you send reports to clients, send them as your own work: your name, your layout, a link they can open without an account. And if the last line of the report is a decision, next month’s report already has its first question.