How to have your members checked every Monday without asking

Who this is for: heads of membership and anyone responsible for retention. What you'll need: Claude connected to your membership data, which your IT lead sets up once. How long: thirty minutes to set up, then it runs itself. What you'll have: a short list of members worth a phone call, in your inbox, every week.

The difference between a report and a routine

Most retention reporting is annual, because that's how often somebody has time to build it. By then the members it describes have gone.

A routine is the same question asked on a rhythm, automatically, with the answer arriving whether or not anybody remembered to ask. That's a small change and it turns retention from something you review into something you work.

This only works once Claude can actually reach your records rather than a spreadsheet you exported. If that connection is not in place yet, it is an hour of your IT lead's time and it is the prerequisite for everything below.

Step one: decide what "at risk" means before you automate it

Automating a vague question gets you a vague answer every week, forever. Spend twenty minutes on this and the rest is easy.

You're looking for members who have stopped doing things they used to do. Not members who never did them. That's a different and much longer list.

1. Pick three or four signals you actually hold: portal sign-ins, CPD entries, event bookings, an unanswered enquiry, a failed payment that was never chased.
2. For each, decide what "stopped" means in months. Somebody who signed in monthly for two years and hasn't for four months is a signal. Somebody who has never signed in is not.
3. Decide the population. Everyone, or one grade, or one branch. Start narrow. A list of 400 names is not actionable and you will stop reading it by week three.
4. Write those decisions down in a sentence or two. You'll paste them straight into the instruction.

Step two: write the standing instruction

Write it as though briefing a new analyst who is competent but doesn't know your organisation. Be specific about the population, the period, the signals and the output.

Ask for a short list, not a report. Twenty names with a reason attached is something a person can act on before lunch. A four-page analysis is something they will read once.

A starting instruction. Change the grades, the months and the cap to yours.

Every Monday, look at members in the Associate and Member grades
who are inside their renewal year.

Find those who have gone quiet: no portal sign-in for 4+ months,
no CPD entry for 6+ months, or no event booking in 12 months
where they had booked at least one in the previous 12.

Return at most 20, most-at-risk first. For each: name, grade,
years of standing, branch, renewal date, and the one sentence
that explains why they are on the list.

Tell me plainly what you could not check.
Compare the count to last week and say if it moved.

That last pair of lines matters more than they look. "What you could not check" catches the week a data source is silently missing. "Compare to last week" is what turns a list into a trend you can feel.

Step three: set it running on a schedule

In Claude, this is a recurring task rather than a conversation you have to remember to start.

1. Open Claude and start a new conversation with your membership connection available.
2. Paste the instruction and run it once, by hand. Do not schedule anything yet.
3. Read what comes back with somebody who knows the members. You are checking whether the names are plausible, not whether the format is pretty.
4. Adjust the thresholds. Too many names means your months are too short; a handful means they're too long. Two or three passes is normal.
5. Once you're happy, set it to run weekly and choose the morning it lands. Monday before ten works because the list gets worked that day or not at all.

Run it by hand at least twice before scheduling it. A routine that was wrong on day one produces wrong output for a quarter, and because it arrives reliably it feels authoritative.

Step four: give the list an owner, or don't bother

This is where these things die. The report arrives, everyone agrees it's interesting, and nobody rings anybody.

Decide before you switch it on: who receives it, what they do with it that week, and what happens to a name once it's been actioned. It doesn't need to be elaborate. A shared list with a column for who called and what happened is enough.

The intervention that works is a person, not a campaign. A membership officer ringing twenty people and asking how they're getting on will save more members than any email sequence, and the list tells them which twenty.

Other routines worth having once this one works

The pattern generalises. Once you trust one, the others take ten minutes each.

• Monthly, during the renewal run: which failed collections are still outstanding, grouped by reason, oldest first.
• Two months before each CPD deadline: who is behind, by category, with how far behind.
• Weekly in the fortnight after an event: which non-member attendees haven't been followed up.
• Quarterly, before the board pack: the retention numbers, cut the same four ways as last quarter, with the gaps declared.

Resist doing all of them at once. Each routine needs an owner, and an organisation that adds five in a month ends up ignoring all five.

What will go wrong, and what it looks like

The list stops changing. Usually means your thresholds are catching a static population rather than movement: people who went quiet long ago and stay quiet. Tighten the "had previously done this" side of each signal.

The list is enormous in month one. Expected. You are seeing the accumulated backlog rather than this week's drift. Work through it once, then the weekly list becomes small and genuinely current.

The same names every week. You need a way to mark somebody as handled, or the routine will keep surfacing a member who has already been rung and told you they're leaving.

And read the "could not check" line every single week. The most likely failure is not a wrong answer. It's a data source quietly dropping out while the report keeps arriving, looking exactly as confident as it did before.

Do this yourself, or have it done

The instruction, the thresholds and the schedule are yours, and they should be. The judgement about what counts as at-risk is membership knowledge, not technical work. You'll write a better one than we would.

Where we help is what sits underneath: making sure the signals you want to watch are actually being recorded, agreeing definitions once so the routine and your board pack don't disagree, and turning a routine you've come to rely on into something with a proper owner, a record of what it found and what was done. That's Maintain and Evolve. If you already have a spreadsheet where somebody tracks at-risk members by hand, bring it. It encodes years of judgement nobody has written down, and it's a far better starting point than a blank page.