Building a collections process that scales

Start with one owner per account

Every customer account has exactly one named person accountable for cash from it. Not a department, not a shared inbox.

In a group the natural owner is whoever the customer already knows, usually the entity's account manager. Central finance runs the cadence and holds the queue; the owner carries the outcome and makes the difficult call. Where several entities bill the same customer, appoint one group-level owner even though the invoices sit in different ledgers.

Two rules make it real: the owner's name sits on the account record, and unclaimed accounts default to whoever signed the contract. Ambiguity is where receivables age.

The cadence before the due date

Most of the value in collections happens before anything is late, and it costs almost nothing.

Three light touches. At invoice issue, confirm it reached the payables contact with the reference or PO number their system requires. Seven to ten days out, on invoices above a threshold you set, confirm it is approved and scheduled. Two days out, a short reminder of amount and date.

This is not chasing, it is an approval check. A missing PO number found on day minus eight is a five-minute fix; found on day plus forty it is a month of lost cash.

The cadence after the due date

After the due date the cadence tightens and the channel escalates, on a schedule you set once.

Day plus one, automated reminder: factual, no apology. Day plus seven, personal email from the owner asking for a payment date rather than for payment. Day plus fourteen, telephone call, the first point at which you reliably learn the real reason for non-payment. Day plus twenty-one, formal notice restating terms and amount, copied to the owner's counterpart. Day plus thirty, escalation.

The exact days matter less than two rules. The channel has to change, because a fourth email carries less weight than a first phone call. And every step needs a fixed date, so nothing waits on someone's judgement.

The escalation ladder

Escalation means the seniority of the conversation rises on a schedule, and it should read to the customer as process, not mood.

Rung one is the credit controller running cadence. Rung two is the account owner, in a commercial conversation. Rung three is the entity's general manager, discussing terms, holds, and payment plans. Rung four is the group owner, speaking to their counterpart. Rung five is external, a decision to write off a relationship.

Trigger each rung on days past due and amount together, never age alone, and record the current rung on the account. A customer who reaches rung three twice in a year is telling you something about your terms, not your collections.

When to involve the relationship owner

The relationship owner comes in when the reason for non-payment stops being administrative.

A missing reference or a wrong contact is clerical and should never reach them. A dispute, an approval block that has survived two cycles, or a stated inability to pay all need someone who can trade something: a credit, a revised schedule, a decision to keep supplying. Earlier wastes senior time; later means the first serious conversation happens after the goodwill is gone.

Brief them in a paragraph: what is owed, what was promised and broken, what outcome you want.

Documenting promises to pay

A promise to pay is a record with three fields: amount, date, and the person who made it. Anything less is a feeling.

Log it against the invoice as it is made, and schedule the check for the promised date rather than the next cadence step. A kept promise resets the account to normal cadence. A broken one skips a rung, because it is the strongest early signal of eventual bad debt. Atradius, in its 2025 North America B2B Payment Practices Barometer, puts around 6% of long-outstanding invoices as eventual bad debts, and the accounts that get there almost always broke two or three promises first.

Credit terms, limits, and holds

Collections cannot fix a credit decision that was wrong at the point of sale, so the process has to reach back into terms and limits.

Three mechanisms. A credit limit per customer, set at group level where several entities sell to the same counterparty. Terms that reflect risk rather than habit, so new customers start shorter and earn longer. And a supply hold that triggers at a defined threshold, with one named person authorised to release it.

The hold is the part groups avoid and the only lever with real force. It works only if defined in advance and applied consistently: a hold waived on request is not a policy. Decide the threshold when nobody is angry. The Federal Reserve's 2024 Small Business Credit Survey of 7,653 employer firms found 51% reporting uneven cash flows: set limits before you need them.

The arithmetic of buying cash expensively

Offering 2/10 net 30 costs roughly 36 to 37% annualised, which makes it one of the most expensive forms of financing available to you.

You give up 2% to receive 98% twenty days early: 2/98 = 2.04% for twenty days, and 365/20 = 18.25 such periods a year, so 2.04% x 18.25 = 37.2%. Arithmetic, not a study.

If your borrowing cost is 10%, that pays nearly four times the market rate for cash you were owed anyway. On $2,000,000 of annual credit sales, taken up by half your customers, it costs $20,000 of margin every year.

Two defensible reasons to offer it: you are genuinely liquidity constrained, or it replaces collections cost and bad debt in a high-risk segment. Neither justifies a discount left on the template because it was there when you arrived.

How the process changes across a group

The process is identical per entity. What changes is that the customer, the credit limit, and the escalation are group-level facts while the invoices are entity-level records. Four rules cover it: one relationship owner per customer, one group credit limit consumed by whichever entities sell, one escalation ladder so nobody receives three unrelated demand letters in a week, and one consolidated exposure figure visible to whoever approves the next order.

Take a group of eight entities, four of which invoice the same regional distributor, under one group owner and a $140,000 group limit. Entity A is owed $18,500 on net 30, four days past due; Entity B, $64,000 on net 45, sixteen days past due; Entity C, $9,200 on net 30, fifty-eight days past due; Entity D, $37,300 on net 60, not yet due. Exposure is $129,000.

Now watch the cadence. Entity A has had the day-one reminder and nothing more. Entity B is at the telephone rung, and the call yields a promise to pay on the twenty-eighth, logged with the amount and the name behind it. Entity C has broken two promises, so it skips a rung to the entity general manager, who can offer an instalment schedule.

Then the hold, the one genuinely group-level decision. The distributor wants Entity D to release another $22,000, taking exposure to $151,000, past the limit. One named person decides against the group figure and the promise log: hold until Entity B's $64,000 clears, leaving $87,000 with the new order included, then release. Without the group figure and the promise log, the $22,000 ships.

Can QuickBooks or Xero do this?

Both handle the single-entity mechanics well. Neither supports the group-level parts.

They run the reminder cadence, produce the aging report that feeds the queue, issue statements and hold terms per customer. For one entity, most of the cadence above runs on tools you already pay for.

They do not hold an account owner, a reason code, or a promise-to-pay record as structured data, so those live in a spreadsheet. They do not enforce a credit limit across separate company files. And they will not show one customer's total exposure across entities, which is the input the hold depends on. Why that gap exists is covered in why aging reports alone won't fix collections.

Where to start

Write the cadence on one page: owners, days, channels, escalation triggers, hold threshold. Run it against your ten largest overdue balances for a month before buying anything.

Then check whether your receivables data is current enough to run a process on at all: if the books close late, every step above starts from stale information. Current books and a fast close on top of QuickBooks Online or Xero is what cruisr does today. Get in touch if that is your constraint, or read the 13-week cash flow forecast for what to do with reliable data once you have it.