How to Build a Dunning Sequence That Actually Collects on Time
Every B2B finance team has the same drawer of dunning emails: three templates, one for polite, one for firm, one for the lawyer letter. They get pasted out of a Google Doc, sent from a shared inbox, and mostly ignored. Then someone rewrites them once a year and nothing changes.
A dunning sequence that actually collects is a different object. It starts before the invoice is late, it branches on customer behavior, and it produces dated commitments, not conversations.
What is a dunning sequence and why do most of them fail?
A dunning sequence is the ordered set of reminders, calls, and escalations you send to a customer between the moment you issue an invoice and the moment they pay it. In a B2B context, it usually spans 45 to 90 days and touches AP contacts, the buyer, and the account manager, in that order of intensity.
Most sequences fail for three specific reasons.
- They start too late. The first touch fires on day 1 past due, which is often already inside the customer's internal payment cycle. By then, the invoice has either been approved and scheduled, or it has been rejected for a reason nobody told you about.
- They send the same cadence to everyone. A customer that pays like clockwork at day 32 gets the same firm day-15 nudge as a customer that has never paid inside 70 days. The reliable payer notices. The unreliable one does not care.
- They lose promises to pay. Someone on the customer's AP team says "we'll pay Friday" on a Tuesday call. Nobody writes it down against the invoice. Friday comes and goes. The next touch fires two weeks later as if the promise never happened.
Fix those three, and the volume of dunning your team sends goes down while collections go up.
What are the 5 touches in a working B2B dunning cadence?
Not more. Not fewer. Five touches across roughly 60 days, with a defined channel and owner for each.
| Touch | Timing | Channel | Owner | Purpose |
|---|---|---|---|---|
| 1 | 5 to 7 days pre-due | AR clerk | Confirm invoice received, PO valid, no dispute | |
| 2 | Day 1 past due | AR clerk | Neutral reminder, payment link, wire details | |
| 3 | Day 15 past due | Email plus voicemail | AR lead | Firmer tone, request a dated commitment |
| 4 | Day 30 past due | Phone plus written notice | AR lead | Live conversation, escalate to buyer if AP unresponsive |
| 5 | Day 45 to 60 past due | Escalation | Account manager or CFO | Service hold, credit hold, or account-level call |
The pre-due touch is the one most teams skip and the one with the highest yield. Every hour spent confirming the invoice landed is an hour not spent chasing a payment that was blocked on a missing PO.
How should tone shift as the invoice ages?
Tone in dunning is not about being nicer or meaner. It is about matching the register your customer's AP team responds to at each stage.
- Pre-due and day 1. Assume competence, not delay. "Just confirming this landed" reads like a workflow check, not a chase. Reserve exclamation marks and "please" pileups for later.
- Day 15. Ask directly for a payment date. Not "any update" but "can you confirm the day this will remit". A vague ask gets a vague answer.
- Day 30. Name the risk in writing. "This invoice is 30 days past due. If we do not receive a payment date by end of week, we will need to escalate to your account owner." The customer needs to know the next lever is real.
- Day 45 plus. Escalate to a human on your side, not just a firmer email. A CFO-to-CFO call resolves more disputes than a fifth reminder ever will.
The cadence is procedural. The tone is behavioral. Both have to shift, and neither can shift without the other.
Who owns each step of the sequence?
Ownership decays if it is not named on the row. Every touch needs a specific person, not a queue.
- AR clerk. Owns touches 1 and 2. Handles the volume, keeps the log, flags the invoices where automation stops and judgment starts.
- AR lead or controller. Owns touches 3 and 4. Makes the calls, negotiates payment plans, decides when an invoice moves to the dispute queue.
- Account manager. Owns touch 5 for accounts where the commercial relationship matters more than the invoice. Loops in customer success only when there is a genuine service issue.
- CFO or VP finance. Owns the exception path. When a $200K invoice hits day 60, the call goes from your CFO to theirs, not from your AR lead to their AP clerk.
If any of these roles is blank, the invoice ages into the next bracket by default. That is not a discipline problem, that is a design problem.
How do you branch the sequence by customer behavior?
The single biggest lever in dunning is not the cadence itself. It is matching the cadence to how each account actually pays.
Four segments, each with a different sequence.
- Reliable, under 45 days average. One pre-due confirmation and one day-1 reminder. That is it. Do not send touch 3 unless payment slips past day 30.
- Slow but predictable, 45 to 70 days. Full 5-touch sequence, but move touches 3 and 4 later by 10 days each. They will pay, just later than terms.
- Erratic, big variance. Start earlier, at 10 days pre-due. Use touch 3 by day 10, not day 15. These accounts need pressure to convert to a promise to pay early in the cycle.
- Chronic late, above 80 days. Escalate to the account manager at touch 3, not touch 5. Consider prepay or credit holds on new orders. The sequence is not the solution here, the commercial relationship is.
The mistake is running segment 4's cadence against segment 1. That is how you lose a $500K reliable account trying to save $8K of interest on a paid invoice.
What data do you need to run a dunning sequence?
The minimum useful record per invoice.
- Customer name, contact, and email
- Invoice number, amount, and issue date
- Due date and cancellation-free grace window
- Days past due, refreshed daily
- Account segment based on last 12 months of payment history
- Current touch, next touch date, and owner
- Promises to pay: date made, amount, promised remit date, status
- Disputes: date raised, reason, owner, status
- Communication log with timestamps, threaded to the invoice
You cannot run this in email plus a spreadsheet past about 200 open invoices. The bookkeeping cost of keeping segment, touch state, and promise status current becomes higher than the cash you are chasing.
The mistake to avoid
Most AR teams treat dunning as a volume problem. Send more, send firmer, send sooner. The teams that actually move DSO treat it as a routing problem: get every invoice to a dated commitment, then hold that commitment. Five touches, four segments, one named owner per step, and a promise-to-pay log that survives when the AR person is out. Everything else is decoration.
Frequently asked questions
When should the first dunning touch go out?
Before the due date, not after. A soft confirmation 5 to 7 days ahead of due catches missing invoices, wrong PO numbers, and approval bottlenecks while there is still time to fix them. Teams that only start after the invoice ages tend to see 30 to 40% of their late payments trace back to something the customer could have flagged pre-due.
How many touches should a dunning sequence have?
Five is the working answer for most B2B contexts. Pre-due confirmation, day 1 reminder, day 15 firmer reminder, day 30 phone call plus written notice, and day 45 to 60 escalation to the account manager or a hold on services. More touches inside 60 days annoy good customers. Fewer touches let late payers stall.
Should reminders come from a person or a shared inbox?
From a named person on your AR team, replying-to a shared inbox. Personal from-lines get roughly double the reply rate of a generic accounts@ address, but the shared inbox lets the whole team see the thread and cover if the AR owner is out. Automation should make the sender look human, not the opposite.
What is a promise to pay and how should you track it?
A promise to pay is a dated commitment from the customer, in writing or on a logged call, to remit a specific amount by a specific day. Track it on the invoice, not on the account, and follow up the morning after the promised date if payment has not landed. Promise-to-pay accuracy is one of the strongest leading indicators of true bad debt risk.
How do you know a dunning sequence is working?
Three metrics move in the right direction within 60 days: median days to pay tightens by 5 to 10 days, promise-to-pay kept rate climbs above 75%, and the share of AR over 60 days drops. If touches go up but those numbers do not move, the cadence is not the problem. Segmentation, tone, or ownership is.
Turn your AR into a cash forecast
Melenyn syncs open invoices from NetSuite, QuickBooks, or Xero, runs adaptive dunning sequences, and tells you which week the money actually lands.
Request early access