Ordered from what you can start on Monday with no budget and no permission, to what needs a system change. Nothing here involves being harsher with borrowers — collections pressure is the least effective lever on this list and the most expensive one, because it costs you the repeat customer who was going to pay anyway.
If PAR 30 is not yet a number you trust, start with how it is actually calculated — several of the ways it gets flattered will make the list below look like it is working when it is not.
Free, this week
1. Call on day three, not day thirty
The single highest-return change most lenders can make. A borrower three days late has usually forgotten, been paid late themselves, or had a bad week; a borrower thirty days late has made a decision about you. The first conversation is a reminder and costs nothing. The second is a negotiation.
Set the trigger at one to three days past due, make the first contact a friendly one, and measure how many of those never become arrears at all.
2. Put the loan account number in every SMS
A large share of unmatched paybill payments are unmatched because the borrower typed something else in the reference field — usually their ID, because that is the number they know. Printing the exact string they should type, in every reminder message and on every statement, moves the match rate more than any matching logic will.
This is cheap, immediate, and it also shortens your suspense list. See M-PESA paybill reconciliation.
3. Move due dates to the day after people are paid
A due date on the 1st competes with rent. A due date on the 28th of a salaried borrower's month competes with nothing. For traders, align to their market days rather than to the calendar. This is a product configuration, it costs nothing, and its effect on on-time collection is larger than anything you will do downstream.
4. Give the officer their own portfolio, visibly
An officer who can see their own PAR, updated daily, next to their colleagues' behaves differently from one who hears about it monthly in a meeting. This is not about targets or bonuses — it is about the number being visible to the person who can move it, on the day they can move it.
5. Stop counting a promise to pay as a collection
A promise-to-pay is worth recording — it is real information and a broken one is a strong signal. It is not a payment, and any collections dashboard that shows PTPs in the same column as cash is telling the team a comfortable story. Track PTP kept-rate as its own number; it is one of the best predictors you have.
Cheap, this month
6. Split "in arrears" into the three reasons
Late-and-will-pay, struggling-but-engaged, and gone. These need completely different treatment and most collections processes apply the same one to all three — which wastes effort on the first group and loses the second.
You do not need software for this. A field with three values, filled in by the officer after the first call, will reshape how the week is spent.
7. Escalate on a schedule, and write the schedule down
Day 3 SMS, day 7 call, day 14 field visit, day 30 supervisor call, day 45 formal demand. The exact steps matter less than that they are the same for everybody and that skipping one is visible. Ad-hoc collections is indistinguishable, from the outside, from favouritism.
8. Look at first-payment default separately
A borrower who misses their first instalment is an origination problem, not a collections problem, and treating it as the latter fixes nothing. Report first-payment default by officer and by branch. A single officer with a high rate is a training or an integrity question; a whole branch with a high rate is a product or a market question.
9. Check your restructuring, honestly
Restructuring a genuinely distressed borrower who will now pay is good business. Restructuring to reset the days-past-due counter is hiding a loss and delaying it. Count restructures monthly, report them next to PAR, and look at what proportion of restructured loans go back into arrears within two cycles. If it is high, the restructures are cosmetic.
Structural, this quarter
10. Fix the reference-number convention itself
If your loan numbers are long, mixed alphanumeric, or change between documents, borrowers will keep typing something else. Short, numeric, unambiguous, and — ideally — stable across a customer's loans, so a repeat borrower does not have to learn a new one. This is a schema decision with a direct effect on collections, which is not obvious until you have watched a suspense list grow.
11. Automate the arrears run, and post penalties with journals
Arrears status calculated by a scheduled job at a fixed time each day is a number everyone can rely on. Arrears calculated by whoever ran the report is a number that changes depending on who you ask, and it makes every conversation about collections into a conversation about the data.
While you are there: post the penalty to the ledger when you accrue it, or do not accrue it. Doing one without the other is item 2 in why your loan book and your ledger disagree.
12. Score by cohort, and stop lending into the bad one
Group your disbursements by month, product, branch and officer, and look at PAR 30 for each cohort at the same age — three months after disbursement, say. This removes the growth effect that makes a fast-growing book look healthy, and it usually identifies one product or one branch responsible for a disproportionate share of the problem.
Most PAR problems, examined this way, turn out to be concentrated rather than general. That is good news: a concentrated problem can be fixed by changing one thing, and a general one cannot.
Two things deliberately not on this list
Harsher collections. Beyond the point of consistent, prompt, professional contact, additional pressure produces very little additional money and a great deal of reputational damage — and conduct in debt collection is now explicitly regulated for digital credit providers. The borrower you harass into paying does not come back for the loan that would have been profitable.
Raising interest rates to cover losses. It selects for borrowers who are less price-sensitive because they are less likely to repay. The arithmetic looks fine for one quarter and then does not.