Buyers almost never stop paying suddenly. The pattern is consistent: payment dates start slipping, amounts start arriving short, reminders stop getting replies, and only then does a full instalment get missed. Every step of that is already recorded in your own system, which means default is predictable two to four months ahead without any new data collection.
The six signals
| Signal | What to measure | Why it predicts |
|---|---|---|
| Date drift | Average days late, this quarter versus last | Drift precedes default more reliably than any single late payment |
| Partial payments | Count of instalments settled in more than one receipt | Indicates the instalment size no longer fits the buyer's cash flow |
| Reminder engagement | Replies and payments within 72 hours of a reminder | Disengagement is the strongest single signal |
| Channel switching | Bank to MFS, or larger to smaller amounts | Often signals a change in who is paying or how |
| Contact reachability | Calls answered, numbers changed | An unreachable buyer is already a recovery case |
| Balance trajectory | Outstanding as a share of price versus expected at this stage | Catches buyers slipping quietly across a long plan |
None of these requires new data. All six are already in the receipt history, the message log and the call log — assuming those live in the same place as the schedule.
Building a score you can explain
Keep it simple enough to defend:
- Date drift over 15 days: 3 points.
- Two or more partial payments in the last quarter: 2 points.
- No reply to the last two reminders: 3 points.
- Number unreachable in the last 30 days: 2 points.
- Outstanding more than 10 percent above the expected curve: 2 points.
Six or more points puts a buyer on the priority list. Ten or more means a manager calls this week. The specific weights matter far less than applying them consistently and reviewing them against outcomes after a quarter.
What to do with the list
The point of a risk score is to change who gets called first, not to change how they are spoken to. A high-risk buyer at month three still has options: a restructure from the outstanding balance, a shorter payment holiday, a revised schedule that matches their actual income pattern. Those conversations work when the balance is still manageable and fail when it is not. The mechanics of doing them properly are in the overdue recovery playbook.
The project-level signal nobody watches
If risk scores rise across a whole project rather than among individual buyers, the cause is usually the site. Buyers who drive past a stalled construction site stop paying, and no amount of individual follow-up fixes that. The response is progress communication and, where the plan allows, milestone-linked billing that only charges for work actually done. See milestone billing and what a delay costs.
Measuring whether the score works
After one quarter, compare: what share of buyers who actually went 90 days overdue were flagged at least two months earlier, and what share of flagged buyers turned out fine. The first number is the value; the second is the cost. A score that flags forty percent of your book is not a score, it is an alarm nobody will listen to. Tune until the flagged list is small enough for someone to actually call.
What to do next
Take the buyers who went 90 days overdue in the last year and look at their payment history in the six months before. If the drift and partial payments were visible — and they usually are — the same pattern is visible right now in buyers who have not defaulted yet — see risk built from the payment history.