Collections Craft: five levers that move your days past due
Collections is not about making more calls. Five levers (prioritization, channel, cadence, promises, escalation) with worked numbers, and a 30-minute weekly huddle to run them.
Your team has capacity for 240 touches this week and needs 385. Collections is not about making more calls; it is about choosing which 240, how to make them, when, and what rule sits behind each choice.
The book we will use all week
Every number below comes from one illustrative book, so the five levers add up to one story. Credit sales are 10.0m in a 30-day month. Ending AR is 11.0m, of which 8.0m is current and 3.0m is past due. None of this is real data; it is a worked example you can rebuild with your own figures.
| Measure | Formula | Result |
|---|---|---|
| DSO | 11.0 ÷ 10.0 × 30 | 33 days |
| BPDSO (best-possible DSO) | 8.0 ÷ 10.0 × 30 | 24 days |
| ADD (Average Days Delinquent) | DSO − BPDSO | 9 days |
Here is the catch. Collections does not own DSO. It owns the part of ADD that sits in past-due invoices it is allowed to chase. Each 1.0m past due is worth 3 days of ADD. If 0.8m of the 3.0m is disputed or short-paid, and owned by someone else, that is 2.4 days collections cannot move. The other 2.2m is 6.6 days it can.
That is the job in one number. Collections turns past-due invoices into promises, and promises into cash, once terms have run out. Most teams stumble on the same three things: the wrong list (sorted by age), the wrong channel (email for everyone) and the wrong time (set by the calendar, not by the customer’s payment run). Promises go untracked, and escalation follows mood.
Here are the five levers, in the order you would pull them.
Lever 1: Prioritization. The oldest invoice is rarely the one to chase first
A past-due list is a queue, and collector time is the constraint. Oldest-first feels fair, but it spends the first calls on the accounts least likely to move. Value-first is easy to defend, but it spends them on the biggest balances whether or not a call changes anything. Neither asks the useful question: what will this touch change?
Rank by expected recovery instead.
Expected recovery of a touch = uplift in P(pay) × balance − cost of the touch
Uplift is the extra chance of paying within a window, say 14 days, that the touch creates. A customer who would have paid anyway earns the touch no credit. In the illustrative book, a call costs 12.
| Account | Days past due | Balance | Uplift | Expected value |
|---|---|---|---|---|
| A | 110 | 25,000 | 2 pts | 488 |
| B | 90 | 150,000 | 4 pts | 5,988 |
| C | 60 | 80,000 | 20 pts | 15,988 |
| D | 45 | 200,000 | 2 pts | 3,988 |
| E | 30 | 120,000 | 25 pts | 29,988 |
| F | 20 | 40,000 | 30 pts | 11,988 |
Account A has the oldest invoice, and it ranks last. Account E is only 30 days late, and it ranks first. With capacity for three touches today, the rule you pick decides what you earn:
| Rule | First three calls | Value of those touches |
|---|---|---|
| Oldest-first | A, B, C | 22,464 |
| Value-first | D, B, E | 39,964 |
| Expected recovery | E, C, F | 57,964 |
To make it work: segment risk and value into a 2 × 2 and name a treatment per cell, let capacity set the cutline, keep an age backstop so nothing sits untouched beyond N days, and estimate uplift from held-out accounts, not from who paid after a call. Re-rank daily, because balances, promises and risk all drift.
And the weekly capacity? Four collectors at 60 touches each is 240 touches against 385 needed, so 62.3% is covered. Ranking decides which 62.3%.
Lever 2: Channel. The cheapest contact is not the cheapest collection
Once you know who to chase, decide how: phone, email, a message in the customer’s AP portal, a letter, short messaging (only where the customer has agreed), or a visit for key accounts. Judge each channel on what it collects, not on what it costs to use.
Contact-to-promise rate = promises obtained ÷ contacts made
Cost per kept promise = (contacts × cost per contact) ÷ promises kept
| Channel | Contacts | Cost per contact | Promises | Kept | Contact-to-promise | Cost per kept promise |
|---|---|---|---|---|---|---|
| 1,000 | 1 | 40 | 24 | 4.0% | 41.7 | |
| Portal | 300 | 2 | 45 | 36 | 15.0% | 16.7 |
| Letter | 200 | 4 | 10 | 6 | 5.0% | 133.3 |
| Phone | 250 | 12 | 100 | 80 | 40.0% | 37.5 |
Email is the cheapest channel per contact, at 1. Yet phone beats it per kept promise, and the portal beats both. A cheap channel that rarely produces a promise is expensive by the only measure that matters.
What to do with it: write a channel matrix by segment and balance. Phone and visits suit large or risky accounts; portal and email suit small, reliable ones; messaging only with agreement. Log every contact (channel, person, outcome, next step) in one place so channels can be compared. Change one channel at a time and compare against a similar group. Pair cost per kept promise with the value kept, because small promises are cheap.
Respect consent, working hours and data-protection rules, and check your legal and compliance requirements.
Lever 3: Cadence. When a touch lands beats how it sounds
A cadence is a dunning sequence with fixed offsets from the due date: a pre-due touch, then touches after. Before you rewrite a single template, read your cure rate by touch.
Cure rate at touch n = invoices paid after touch n ÷ invoices reaching touch n
Illustrative cohort of 200 invoices:
| Touch | Invoices reaching it | Paid after it | Cure rate |
|---|---|---|---|
| 1 | 200 | 70 | 35.0% |
| 2 | 130 | 39 | 30.0% |
| 3 | 91 | 18 | 19.8% |
Cumulatively, (70 + 39 + 18) ÷ 200 = 63.5% of the cohort cures. The other 73 invoices (36.5%) go on to escalation. Each touch earns less than the one before, so a touch whose cure rate is close to zero should go.
A pre-due touch earns its place too. If on-time payment rises from 60% to 72% and late invoices average 15 days late, that is (72% − 60%) × 15 = 1.8 days saved. At daily credit sales of 0.33m (10.0m ÷ 30), that releases about 0.6m of cash.
Then there is the payment-run trap. Say a customer pays every Thursday. An invoice due on Friday waits 6 days for the next run unless it is approved before the cut-off. A touch that lands after the cut-off is a touch wasted. So land touches before the customer’s payment-run cut-off and ahead of month-end, not after.
The order of tests matters: timing first, channel second, wording last. Fix the offsets and hold them, so customers learn when you really chase. Make each touch add information or a deadline instead of repeating the last one. Hold a no-touch control group for a quarter if you need to prove the value of the pre-due touch.
Lever 4: Promises. A high kept rate can hide the big broken ones
A promise has a date, an amount and a named payer, logged in one place. Anything else is a hope. The number that flatters you most is the kept rate by count, so always set it next to the kept rate by value.
| Measure | Calculation | Result |
|---|---|---|
| Kept rate by count | 40 paid by the promised date ÷ 50 promises made | 80.0% |
| Promises kept by value | 1.116m paid by the promised date ÷ 1.8m promised | 62.0% |
| Value to re-chase | 1.8m − 1.116m | 0.684m |
| Promise coverage | 1.8m under promise ÷ 3.0m past due | 60.0% |
An 80% kept rate sounds healthy. A 62% kept rate by value says the large promises are the ones that broke, and 0.684m still needs chasing. The 18-point gap between the two is the story.
The routine around the number is short:
- Schedule the follow-up for the promise date when the promise is made, not a week later. Confirm payment is initiated and ask for the remittance.
- Count a partial payment as kept at its value, not as a full promise.
- On a broken promise, re-contact within one business day and find the reason.
- Accept one re-promise. The second break goes to escalation.
- Review broken promises above a value threshold within one business day.
Track the gap between count and value, promise coverage of past-due value, promises renewed more than once, and the hours from a broken promise to re-contact.
Lever 5: Escalation. A written rule, not a mood
Escalating fast feels serious. Escalating a disputed invoice just burns the relationship and does not release the cash. So the first step is a check: is the dispute flag set? If yes, route the invoice to its owner. If no, escalate by rule.
Escalate when days past due ≥ trigger(segment) AND balance ≥ threshold(segment)
Write the triggers down by segment before anyone is annoyed, and publish the matrix to Sales and Collections. Illustrative tiers:
| Tier | Who steps in | Days past due | Balance |
|---|---|---|---|
| 1 | Team lead | 15 or more | 10,000 or more |
| 2 | Head of credit, with the account owner | 30 or more | 25,000 or more |
| 3 | Executive sponsor | 45 or more | 50,000 or more |
A 30,000 invoice that is 32 days past due goes to Tier 2, not Tier 3, because 32 is under 45. Every tier names who steps in, the message (the days, the amount, the next step) and the exit ramp: a dated plan, a part-payment or a named contact. Key accounts get an executive sponsor, and the escalation comes with a business case.
Three numbers tell you whether the rule works:
| Number | Calculation | Result |
|---|---|---|
| Cost of delay | 30,000 × (10% ÷ 365) × 32 days, at 10% a year | about 263 |
| Escalation rate | 18 escalated ÷ 240 past-due invoices | 7.5% |
| Reclassified as disputes | 4 of the 18 escalations | 22.2% |
That last row is the warning light. If one in five escalations turns out to be a dispute, the check at the start of the flow is not being run.
The weekly huddle: five numbers, 30 minutes
The five levers only stick if one meeting reviews them. The format is simple: the same agenda every week, numbers first and stories second, one action each, so the trend becomes the conversation.
CEI = (beginning AR + credit sales − ending total AR)
÷ (beginning AR + credit sales − ending current AR) × 100
On the illustrative book from the start of this article:
| Number | Calculation | Result |
|---|---|---|
| CEI | (10.0 + 10.0 − 11.0) ÷ (10.0 + 10.0 − 8.0) = 9.0 ÷ 12.0 | 75.0% |
| ADD | DSO 33 − BPDSO 24 | 9 days |
| Past due over 60 days, share of AR | 1.1 ÷ 11.0 | 10.0% |
| Promises kept by value | 1.116m ÷ 1.8m | 62.0% |
| Contact coverage of past-due accounts | 90 accounts touched ÷ 120 needing a touch | 75.0% |
There is a sixth number worth knowing even if it is not on the agenda: the touch gap. Touches needed minus touches available is 385 − 240 = 145 a week, or about 2.4 collectors at 60 touches each. Ranking decides which touches happen; the gap decides how many you can afford to skip.
Here is the plan for Monday, in order:
- Prioritize. Rank the worklist by expected recovery, with an age backstop.
- Channel. Set a channel per segment and log every contact.
- Cadence. Fix the offsets and add a pre-due touch for accounts that slip.
- Promises. Define them, follow up on the date and report by value.
- Escalation. Publish the matrix and check for disputes first.
Try it on your own book
You do not need new tooling to start. Pull last month’s figures and answer these in order:
- What is your ADD, and how much of it is past-due value that collections is actually allowed to chase?
- Does your worklist sort by age, by value, or by what a touch changes?
- What is your cost per kept promise by channel, not just your cost per contact?
- Which touch in your sequence has the lowest cure rate, and does it still earn its cost?
- Do you report promises kept by count, by value, or both?
- Is the dispute flag checked before the executive call?
- What is on your weekly collections agenda today, and what should be?
The myth that sits behind most weak collections performance is “more calls mean more cash.” It does not. Better ranking, better timing and better follow-through do.
All worked numbers in this article are illustrative. Use the formulas with your own data, vendor-neutral and tool-agnostic, and keep your legal and compliance rules in view.
- Collections Craft: five levers that move your days past due
- The collector who wins every segment and still loses
