How a dispute decision gets recorded
Individually, a dispute is an annoyance to be cleared. In aggregate, fifteen disputes a month are a description of exactly what is wrong with how the business delivers and bills, written by your customers, for free. This post is about capturing enough on the way past to read that.
Key takeaways
- Resuming the chaser is a decision, not an automatic consequence of closing a dispute.
- A resolved dispute records the reason, the outcome, the elapsed time and who decided.
- Outcome and reason are separate: a quantity dispute can end in a credit note or in a POD.
- The monthly report groups by reason and by customer, which are different questions.
- One recurring reason is a process problem. One recurring customer is a relationship problem.
Resuming the chaser
Closing a dispute does not automatically restart payment reminders, and the distinction matters. A dispute resolved in the customer’s favour ends with a credit note and a new balance, which should be chased on its own schedule. A dispute resolved against the customer ends with an invoice that has now been outstanding for two weeks longer, and immediately firing an overdue reminder at somebody you have just told they were wrong is a poor sequence of events.
- Upheld, credit note raised. The original invoice is settled by the credit note; any remaining balance restarts chasing from the resolution date, not the original due date.
- Not upheld, evidence sent. Chasing resumes, but the first reminder is delayed by a grace period from the sheet — a week is typical — because the customer needs time to act on the reply.
- Partially upheld. The most common outcome in practice. A partial credit note, a new balance, and the grace period applies.
- Withdrawn. The customer says never mind. Chasing resumes at the original schedule, because nothing about the invoice changed.
The record
| Field | Example | Why it is there |
|---|---|---|
invoice | 4412 | What was disputed |
customer | acme-ltd | The other grouping that matters |
reason | quantity | What was claimed |
reason_actual | quantity | What it turned out to be, for measuring the classifier |
outcome | not_upheld | upheld, not_upheld, partial, withdrawn |
value | 0.00 | The credit note amount, if any |
opened | 2026-07-14T09:12:00Z | When the message arrived |
closed | 2026-07-14T11:40:00Z | When a person decided |
decided_by | accounts@example.com | Who |
evidence_gap | none | Or what could not be found, which is the useful one |
reason and reason_actual being separate is what makes the classifier measurable. outcome and reason being separate is what makes the report useful, because “quantity disputes” and “disputes we lost” are different populations and conflating them hides both.
The monthly report
- App integration
- Management
- Analytics
Grouping by reason versus by customer
These answer different questions and both are worth having. Eight quantity disputes in a month, from six different customers, is a delivery process problem — almost certainly split loads being invoiced as one and delivered as two. Fixing it is a change to how deliveries are documented, and it removes eight disputes a month permanently.
Eight disputes in a month from one customer, across four different reasons, is not a process problem at all. It is a relationship that has gone wrong, or a customer with a cash flow problem using disputes to delay payment, and neither of those is fixed by improving delivery notes. Reading a single combined list makes both of these invisible.
Median time, and why median
The mean is useless here because one dispute that took eleven days while somebody was on leave dominates fourteen that took two hours. The median describes the ordinary case, which is what the system was built to improve. Watching it fall from days to hours in the first month is the clearest evidence the thing is working.
What the report deliberately does not include
- Anything per-employee. Who handled a dispute is recorded for audit and never aggregated into a per-person statistic.
- A total value of disputes raised. Only upheld value is real money; disputed value is a number that makes the situation look worse than it is.
- Any customer scoring or risk rating. The system records what happened; deciding what a pattern of disputes means about a customer is a judgement with commercial context in it.
- A forecast. Fifteen a month is not a rate you can extrapolate from, and presenting it as one invites decisions it cannot support.
Next: what all of this costs to run.
All posts