Every BHPH collections operation tracks promises to pay. Very few track whether the promises were kept, and almost none track it by collector. That gap hides more operational dysfunction than any other single reporting hole.
The core problem: PTP volume is trivially gameable
If a collector is measured on promises secured, the optimal strategy is to accept every promise offered — including the ones both parties know aren't real. "Can you do $300 Friday?" "Yeah, probably." Logged as a promise.
That collector posts excellent activity numbers. Their pipeline looks strong. Then Friday arrives and 65% of it doesn't happen, and everyone concludes the customers are unreliable.
Promise volume measures how many conversations ended agreeably. Promise-keep rate measures how many of them were real.
The four reasons promises break
1. The promise was never real (a collector problem)
The customer said something agreeable to end the call. The collector accepted it to close out and move on. No specific amount, no specific date, no confirmation of how the payment would be made.
The tell: vague promises. "Sometime next week." "Around the 15th." "I'll try." A real promise has an amount, a date, and a method.
2. The promise was real but impossible (an underwriting or discovery problem)
The customer meant it and genuinely couldn't do it. Rent came first. The car needed a repair. The hours got cut. The promise was sincere and unachievable at the moment it was made.
The tell: promises that break in the same direction repeatedly for the same customer. This is a signal to restructure, not to keep calling.
3. The promise was real and nobody followed up (a process problem)
This is the most infuriating category because it's entirely self-inflicted. The customer intended to pay, got busy, and nobody reminded them. An unfollowed promise is a note in a file.
The tell: promises that break with no contact between the promise and the due date. Check whether a reminder actually went out. Frequently it didn't.
4. The promise was kept and the system didn't record it (a systems problem)
Rarer, and worst of all. The customer paid, the payment failed to post, and your system now shows a broken promise on a customer who did exactly what they said. The next call accuses a paying customer of being delinquent, and you've damaged a good relationship with your own reporting error.
The tell: customer disputes about payments they insist they made. Take those seriously — they're often right.
Measuring it properly
Promise-keep rate = promises kept (on date, in full) ÷ promises made.
Track three variants, because they answer different questions:
| Variant | Definition | What it tells you |
|---|---|---|
| Strict | Full amount, on the promised date | Genuine promise quality |
| Practical | Full amount, within 3 days | Real-world recovery |
| Partial | Any payment within 7 days | Whether the conversation moved anything at all |
Then segment by collector, by delinquency bucket, and by how the promise was captured. The collector-level spread is where the actionable information lives — and it will be wider than you expect.
What actually improves keep rates
Require specificity
Amount, date, and method. Not "next week" — "$285 on Friday the 14th, by debit card over the phone." A promise without all three isn't a promise, and shouldn't be logged as one.
Confirm in writing immediately
A text sent while the customer is still on the phone, restating the amount and date. Cheap, fast, and it converts a vague verbal agreement into something concrete that exists outside the customer's memory.
Set the payment up on the call when you can
A promise you can execute now — scheduled debit, autopay enrollment — has a keep rate approaching 100% because there's nothing left to remember. This is the single highest-leverage change available.
Automate the reminder
A reminder the day before the promised date closes the entire category-3 failure mode. This is exactly the repetitive, zero-judgment work that should never depend on a human remembering.
Reject promises you don't believe
Counterintuitive and important. If a customer offers $400 Friday and has broken three consecutive promises, taking that promise is not progress — it's deferring the real conversation by a week. Better: "Let's find a number you can actually hit."
Stop measuring collectors on promise volume
If you measure volume, you get volume. Measure keep rate, or measure dollars actually collected against promises made. The incentive has to point at reality.
The diagnostic that takes an afternoon
- Pull every promise logged last month.
- Mark each kept or broken under the strict definition.
- For 20 broken ones, listen to the original call and sort them into the four categories above.
- Count the distribution.
The distribution is your action plan. Mostly category 1, you have a collector coaching problem. Mostly category 3, you have a follow-up automation problem and it's the easiest one to fix. Mostly category 4, stop everything and audit your payment posting.
Promise-keep rate is one of the nine metrics we'd argue actually predict portfolio health — the rest are here.
See what this looks like on your portfolio
Fifteen minutes on Teams. We'll look at your delinquency, show you real AI text conversations, and tell you honestly whether we can move your numbers.
Book a walkthrough