Playbook · MST Golf
Wrong ICP by every textbook, and the engine still worked
Single-location golf retail. Not a restaurant, not a chain. Filter by industry and this shop is cut immediately. Filter by the shape of the cycle and it is the ideal customer.
- Member base
1,240
Member base
- Active in 30 days
248
Active in 30 days
- Silent past 90 days
585
Silent past 90 days
- Given an action
967
Given an action
The member funnel
57% of the base has already bought again — the question is who is about to stop
detail
41%
Revenue from the top 2% of members
78%
Share of the base given an action
967 of 1,240
The actions returned
Five cohorts, five reasons, and not one of them named in advance
| Cohort | People | Action | Why this cohort gets this action |
|---|---|---|---|
| Champions | 281 | VIP fitting night — first look at new stock | Always full price. They respond to status, not discounts |
| Gold, drifting | 119 | Pull back before the window shuts | High value, and now past their own cycle length |
| Bought the driver, not the balls | 98 | Attach from what usually follows | A pattern that repeats across golf retail |
| Fitted, hasn’t bought | 65 | Paid service → highest-margin product | The hottest lead in the base, and the shortest path to margin |
| Silent past six months | 165 | Reactivate with what they already bought | Value about to leave the base permanently |
- Champions281
- Silent past six months165
- Gold, drifting119
- Bought the driver, not the balls98
- Fitted, hasn’t bought65
- Other cohorts239
- No action this cycle273
The last row is the honest one. 273 members had nothing worth sending this cycle, and the engine said so rather than inventing a campaign for them.
how one cohort is valued
281 × ฿2,800 = ฿786,800
The champion cohort multiplied by its own average basket, for one campaign cycle. That is the size of the opportunity, not a guaranteed result.
the limit of this example
detail
A prototype running on real data
Open the real thing, not a screenshot
detail
https://mst-golf-crm.vercel.app/playbook
The demo runs on a sample dataset. The numbers show the mechanism, not a measured result from live use — measured results come from the pilots, and always ship with a confidence interval.
What this example proves
Size is not the requirement
A base of 1,240 is enough to segment and return real actions. You do not have to wait for tens of thousands.
Industry is not the requirement
Golf retail is not the ICP any playbook would recommend, yet its cycle shape matches everything the engine needs.
The work that disappeared was segmentation
967 people sorted into five cohorts with reasons attached, and nobody did it by hand once.
Wrong ICP by every textbook, and the engine still worked — which is the evidence that our filter should be the shape of the cycle, not the sign above the door.
Want to see this for your own shop?
Send us six months of sales history and we return the first set of next-best actions before you commit.