A propensity-to-pay collections strategy that treats can't-pay and won't-pay customers differently, and forecasts every change before it goes live
More cash, fewer disconnections: every strategy change simulated before it goes live.
See one case, screen by screen ↓Most collections run on one dunning path keyed to arrears age: a reminder at day 30, a letter at day 45, a final notice at day 60 and a field visit at day 70. A customer who forgot to pay and a customer whose hours were cut get the same letters and the same truck. The expensive steps go to the wrong people: a field visit costs $250–$500, and many go to customers who cannot pay to reconnect. Those customers end up disconnected, then written off, and each of those disconnections is a complaint risk and a regulator risk.
Changing the strategy is guesswork. There is no way to see who would move and what it would do to cash, cost to collect, bad debt and complaints before it goes live, so the strategy rarely changes at all. At Lumina Utilities, strategy C-14 has been live since 2 June; under it, 3,870 field visits a quarter go out and $6.58M of bad debt is forecast over the next 12 months.
Estimated split for one change to the collections strategy, from idea to a decision to test, in working days, by hand and with the solution. The 60-day test itself is not counted.
Six agents score, protect, plan, simulate and measure; the collections manager moves the lines and the head of credit approves.
Lumina Utilities serves electricity and water. On Tuesday 13 October 2026 at 08:40, its collections manager opens the arrears book: 148,600 accounts 30 days or more behind, $97.7M. Here is what happens next, screen by screen, in the working solution.
Rosa answers for every account in arrears. Her home screen places each one by ability to pay against willingness to pay, scored at 06:12. Live strategy C-14, live since Tue 2 Jun and approved by Helen Marsh, puts the won't-pay line at willingness 50, the can't-pay line at ability 35, the plan-offer line at 30 and the field-visit floor at $300. That gives 112,751 reminders, 16,781 payment arrangements, 7,492 arrangements with outreach, 11,019 on the disconnection path and 557 held as protected.
“Drag a line on the left, or click a value under a slider. The forecast is worked out against live before anything changes for a customer.”
The overnight run read 148,600 accounts from billing, payments and arrears ageing to 05:40. The ability scorer weighed each bill against estimated income, bureau hardship flags from the file of 11 Oct and assistance enrolment. The willingness scorer read payments, promises and replies to 1.9M contacts in 18 months. The protection check held 557, and the treatment planner booked the week under C-14.
“41,180 texts · 9,620 letters · 3,480 calls · 298 field visits this week, each to a supervisor”
Marisol owes $1,284 on electricity and water, 74 days behind. Her ability score is 41: her bill is about 9% of estimated household income against a Harbourside median of 3%, the bureau has flagged hardship since 18 Aug, and her payments shrank from $148 in June to $45 in August. Her willingness is 36 because three letters since 26 Aug went unanswered. On 28 Jul she called in: “my hours were cut at the warehouse”.
Under C-14: “Final notice sent · field visit booked for Wed 21 Oct”
Rosa clicks 45 under the can't-pay line, and draft C-15 opens against live. 3,339 accounts change treatment: 3,216 move from the disconnection path to arrangement and outreach, and 123 the protection check was already holding move there too. Marisol's dot turns green. The forecast on the right updates before anything goes live.
“If C-15 went live”: field visits 3,870 → 2,786 · disconnections 2,322 → 1,672 · estimated benefit +$2.17M a year
Under C-15 Marisol moves to “Can't pay, not in touch”. The treatment planner proposes an arrangement of $72 a month for 18 months with a referral to the Lumina Bill Help Fund, inside the rules of at most 24 months and no more than 6% of estimated monthly income. Rosa can send the plan by text or make the referral from the same screen.
“Cannot pay now and is not answering letters: a support-team call first, then the plan by text.”
The simulation replays 18 months of treatment outcomes for the draft. Over 90 days: cash collected $53.08M (+$0.18M), cost to collect $1.41M (−$0.27M), disconnections 1,672 (−650) and 533 complaints. Bad debt over the next 12 months falls to $6.21M (−$0.37M). The sensitivity curve shows where the can't-pay line should sit, and where the history is too thin to trust.
“The benefit keeps rising past 45, but so does the band: only 380 won't-pay customers with ability above 45 were offered a plan in 18 months. Test 45 first, then read the result.”
The benefit is built in plain view: cash gained, +$0.72M, and cost to collect saved, +$1.08M, are each the 90-day change counted four times for a year; bad debt avoided, +$0.37M, is the 12-month write-off forecast. The simulator's run lists each step, from the overnight scores to the replay of 212,480 treatments since Apr 2025, and the outcome rates it used for every treatment.
Disconnection path under C-15: $113 cost per account and 38.0 complaints per 1,000, against $21 and 5.0 for arrangement + outreach
Under C-14 the protection check holds 557 accounts off the disconnection path: 118 medical certificate or life support, 196 assistance application pending, 141 age 75+ or registered disability, 64 bereavement or hardship, and 38 disputed bills. Each hold shows its evidence and a re-check date. Edith Carmichael's oxygen concentrator certificate, from her GP practice on 14 Sep, is held to Mar 2027 and cannot be released.
“0 protected customers sent a field visit since 2 Jun”
Rosa stops at 45 and sends C-15 as a challenger: 10% of accounts, 14,860 chosen at random, starting Mon 19 Oct for 60 days, with a forecast of +$2.17M a year if promoted. The approver is Helen Marsh, Head of Credit & Collections.
“The protection check applies to both cells, and every field order still needs a credit supervisor's signature.”
Helen approves C-15 as a 10% challenger, and test T-035 is queued to start Mon 19 Oct on 14,860 accounts. Beside it, T-031 shows how a challenger earns promotion: text first, letter on day 7, lifted payment within 14 days from 57.1% to 61.2% at 97% confidence over 60 days. The test analyst reports; promotion needs Helen.
“Test analyst: the challenger is ahead at 97% confidence over the full 60 days. Promotion needs Helen Marsh.”
The same strategy change, seen by the four people who own it — what their month looked like, and what it looks like now.
Six agents score, protect, plan, simulate and measure; the collections manager moves the lines and the head of credit approves.
Scores each account's ability to pay, 0–100, from the bill against estimated income, bureau hardship flags, payment-size trend and assistance enrolment or likely eligibility, and lists each driver with its direction.
Scores willingness to pay, 0–100, from 24 months of bills paid, promises to pay kept or broken, replies to letters, texts and calls, and portal use.
Before any final notice or field order, checks the account against every protected category — medical or life support, assistance pending, age 75+ or disability, bereavement or hardship, disputed bill — holds it off the field path, records the evidence and sets a re-check date.
Places each account by its two scores against the lines of its strategy, champion or challenger, and books the next action with the right notice clock: reminder, arrangement offer, outreach, final notice or field visit.
Replays 18 months of treatment outcomes to forecast a draft strategy against the live one: 90-day cash collected, cost to collect, complaints and disconnections, and 12-month bad debt, with an 80% band.
Splits accounts into champion and challenger cells at random at the share set, measures the test metric daily and reports uplift and confidence at day 30 and day 60.
Manager sends challengers, head of credit approves, supervisor signs every field order. The agents propose; a named person decides.
Ask in plain words about an account, a line or a test — or say what to change in the strategy. Every answer points to its source.
Why is Marisol Ortega booked for a field visit?
Because C-14 only asks whether she is paying. She has not answered three letters since 26 Aug, so her willingness is 36. But her ability is 41: two short payments since July, a bureau hardship flag from 18 Aug, and a bill near 9% of her estimated income. Under C-14 every non-payer above ability 35 goes to the field. With the can't-pay line at 45 she gets an arrangement instead: $72 a month for 18 months with a Bill Help Fund referral.
What happens if we raise the can't-pay line to 45?
3,339 accounts move from the disconnection path to arrangement + outreach. Over 90 days the simulator forecasts $0.18M more cash, $0.27M less cost to collect, 130 fewer complaints and 650 fewer disconnections; 12-month bad debt falls by $0.37M. That is an estimated $2.2M a year. Above 45 the history is thin, so send it as a challenger first.
Which protected customers would the field list reach this week?
None will be visited. Under C-14, 557 protected accounts sit on the disconnection path and the protection check holds every one: 118 medical or life support, 196 with an assistance application pending, 141 aged 75+ or with a registered disability, 64 bereavement or hardship, 38 disputed bills.
Promote the text-first test for customers who forgot
T-031 is proven: 61.2% paid within 14 days with a text first, against 57.1% with a letter first, at 97% confidence over 60 days. I have sent the promotion to Helen Marsh for approval. The notice clocks and the protection check do not change.
13 screens from the working solution, on its sample data. Pick one to see it large.
The lines, arrangements, notice clocks, protected categories, tests and approvers are settings. Every change is saved as a new strategy version.
| Setting | Default | Choose from |
|---|---|---|
| Can't-pay line | Ability 35 | 25 · 35 · 45 · 55 · 65 on the slider — non-payers below it get a plan and outreach, not the disconnection path |
| Won't-pay line | Willingness 50 | 30 · 40 · 50 · 60 · 70 — below it, a customer is treated as not paying |
| Field-visit floor | $300 | $300 · $500 · $700 · $900 — disconnection-path accounts owing less get letters and calls only |
| Arrangements | Up to 24 months | Smallest instalment $15 a month; largest 6% of estimated monthly income; Bill Help Fund referral with every arrangement + outreach |
| Notice clocks | Final notice day 60 | Reminder day 30 · letter day 45 · final notice day 60 · field visit no earlier than 10 days after · door hanger 48 hours before |
| Protected categories | All on | Medical certificate or life support (always on) · assistance pending · age 75+ or disability · bereavement or hardship · disputed bill · winter protection from 15 Nov |
| Challenger tests | 10% share | Challenger share 5–25%; promote at 95% confidence, day 60 |
| Approvals | Helen Marsh | Send a challenger: collections manager · go live or promote: head of credit & collections · each field order: credit supervisor · release a protection hold: collections manager, with a reason |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = the solution's simulation on the demo book (148,600 accounts in arrears, $97.7M): (+$0.18M cash + $0.27M lower cost to collect) × 4 quarters + $0.37M less 12-month bad debt ≈ $2.17M a year; disconnections 2,322 → 1,672 and complaints 663 → 533 over 90 days; the days split is our estimate · “typical” = published figures (NEADA via WNS, 2025, vendor white paper — one in six US households behind on energy bills; NH PUC staff analysis — electric utility bad debt 0.3–0.9% of revenue; S&C via Ventavid — a field visit costs $250–$500). Lumina Utilities, its streets, accounts, funds and the people named on this page are fictional — characters and sample data in the working solution.
Software that scores every customer in arrears on how likely they are to pay and picks the next action for each account. Propensity-to-Pay Collections Strategy scores two things separately — ability to pay and willingness to pay — so a customer who forgot gets a reminder, one who wants to pay but can't yet gets an arrangement, and one who can't pay gets a plan and outreach instead of the disconnection path.
The ability scorer rates each account 0–100 from the bill against estimated income, bureau hardship flags, the trend in payment size and assistance enrolment. The willingness scorer rates 0–100 from payments, promises kept, replies to contact and portal use. Every score lists its drivers, and no protected characteristic is ever an input.
Yes, twice. The strategy simulator forecasts a draft against the live strategy on 18 months of your own treatment outcomes — cash, cost to collect, bad debt, complaints and disconnections, with a confidence band. Then the change runs as a champion/challenger test on a random share of accounts, and is promoted only at 95% confidence after 60 days.
In the working solution's sample book of 148,600 accounts, moving the can't-pay line from 35 to 45 is forecast to keep an estimated $2.2M a year — more cash, lower cost to collect and $0.37M less bad debt — with 28% fewer disconnections and 130 fewer complaints in a quarter. Those are estimates on demo data; the challenger test measures the real effect on your book.
The protection check holds medical-certificate and life-support, assistance-pending, age 75+ or disability, bereavement or hardship and disputed-bill accounts off the field path before any notice goes out, whatever the strategy lines say. Medical and life-support holds are never released; other holds need a person and a reason.
Yes. It works on combined electricity and water accounts and reads the data a utility collections team already has: billing and payments, arrears ageing, credit bureau hardship flags, assistance enrolment and contact history. Its outputs are a risk segment and next action per account, a prioritisation matrix of debt value by propensity, and champion/challenger results.
Yes. The agents score, check, plan, simulate and measure; they never put a strategy live or sign a field order. The collections manager sends a challenger, the head of credit & collections approves going live and every promotion, and a credit supervisor signs each field disconnection order after a 10-day final notice.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your collections rules, notice clocks, protected categories and a sample of your arrears and treatment history — and it goes live once every quality gate has passed. We will walk you through it on your own arrears book first.
Held bills cleared in two days, not six, with the cause and the fix attached.
1-min demoHigh Bill InvestigationUsage, tariff, weather or leak: the real reason, in plain words.
1-min demoRevenue Assurance AnalyticsRecover the 2-4% that leaks between meter and bill.
1-min demoWe'll run Propensity-to-Pay Collections Strategy on a sample of your own arrears and treatment history.