How long before the numbers mean anything?

Long enough for a typical customer to have had several chances to visit. If your regulars come weekly, six to eight weeks. Reading a cohort at two weeks tells you about signup, not loyalty.

What if my numbers are small?

Small numbers are noisy, so watch direction over several months rather than reacting to a single week. Fifty members moving consistently beats five hundred moving randomly.

Reading your loyalty analytics without fooling yourself

Total signups looks good and tells you almost nothing. What matters is how many people reach a first reward, whether members come in more often, and how each month's new joiners behave afterwards. A rising total can hide everything else getting worse.

The risk with analytics is not having none. It is having a number that goes up and concluding something is working.

The vanity metric

Total signups. It only goes up, it is flattering, and it says nothing about whether the programme changed anybody’s behaviour.

You can double it with a competition and be worse off, because you have added people who do not visit and diluted every rate you calculate.

The metrics that mean something

Share reaching a first reward. Of the people who joined, how many finished? This is the single best test of whether your threshold is realistic.

Visit frequency among members. Are members coming more often than they were? This is what you are actually buying.

Cohort behaviour. Group members by when they joined and follow each group. This is the one that stops a rising total hiding a leaking bucket.

Lapsed regulars. Members who used to come often and recently stopped. The most valuable list in the business and almost nobody looks at it.

Comparing like with like

Averages blend new and old members, so an average always improves when you add signups. Cohorts do not. If you change your threshold, compare the cohorts who joined after against those before, and you have an experiment rather than an impression.

Beware the timezone and the window

Different views answer different questions and legitimately disagree. A daily operational dashboard and a monthly billing count are not the same measurement, and reconciling them by assuming they should match wastes an afternoon.

What to do weekly, and what monthly

Weekly, glance at signups and redemptions to confirm nothing has broken.

Monthly, look at cohorts, completion and lapsed regulars, and make at most one change. Changing two things at once means learning nothing from either.

The honest test

If you cannot say what you would do differently depending on a number, you do not need that number this month.