Credit Limits: The Control That Acts Before the Money Is Gone
Every business that sells on credit eventually discovers its largest debtor is also its favourite customer. A credit limit is the only control that acts before the money is gone — here is exactly what it measures, when it fires, and the one setting that silently switches it off.
Collections is a conversation about money you have already lost control of. A credit limit is the conversation before that — the only point in the cycle where saying "not until the last one is paid" costs you an awkward phone call rather than a receivable. Almost every Kenyan business that has written off a serious debt can name the invoice where a limit would have stopped it, and can also explain why nobody stopped it: the customer was good for it, the order was urgent, and the salesperson had a target.
Which is why the control has to sit in the system rather than in somebody's judgement at five o'clock on a Friday.
Four questions to ask about your own credit control
What is this customer's credit limit?
The usual answer
A number somebody remembers, or "we don't really have one for them".
What it actually tells you
If the limit is not recorded against the customer, it is not a control — it is a preference. A limit that lives in the credit controller's head stops working the day they take leave, which is statistically the day the large order arrives.
What is their exposure right now, including today's unbilled orders?
The usual answer
The balance from the statement.
What it actually tells you
The statement is history. Exposure is the balance plus everything open and unpaid, plus the invoice you are about to raise. A limit checked against the wrong number is a limit set at the wrong level.
Who can override a hold, and is the override recorded?
The usual answer
A manager, verbally.
What it actually tells you
An unrecorded override is indistinguishable from no control at all after the fact. If you cannot list last month's overrides and who granted them, you cannot tell whether the limit is working or being routinely ignored.
When did you last review the limits?
The usual answer
They were set when the customer was onboarded.
What it actually tells you
A limit set three years ago describes a business that no longer exists — theirs and yours. Limits that are never reviewed drift into being either an obstruction on your best customers or a fiction on your worst.
What the check actually measures
The evaluation is more careful than most people expect, and knowing its exact shape is what lets you set a limit that means something. It does not compare the new invoice to the limit. It compares the customer's total exposure after this invoice to the limit.
One customer, one new order
Two details worth noticing. Invoices already settled or cancelled are excluded, so a customer who pays promptly has their headroom back immediately rather than at month end — which is what makes a limit usable rather than an annoyance on your best accounts. And when an existing invoice is being edited, that invoice is excluded from the exposure so it is not counted twice against itself. Both are the sort of thing that quietly decides whether people trust the control or start working around it.
The setting that switches it all off
A credit limit of zero means no enforcement. Not "no credit allowed" — no checking at all. The evaluation returns immediately and the invoice proceeds however large it is.
That default is the right one, because the alternative would be that every business switching this system on discovers it cannot invoice anybody. But it does mean the control is opt-in per customer, and a customer nobody has set a limit for is a customer with unlimited credit. If you take one thing from this piece: the risky account is not the one with a low limit, it is the one with no limit recorded.
How tightly to set a limit
No limit recorded (zero)
The default, and unlimited credit. Appropriate for cash-only customers who will never carry a balance, and dangerous for anybody else precisely because it looks like a blank field rather than a decision.
Two to three months of normal trade
Where most limits should sit for an established account. Loose enough that a good customer never notices it, tight enough that a customer who stops paying hits it within a cycle or two rather than after a year.
One month of normal trade
Right for a newer account or one with a patchy record. Expect to review it upward within a year if they behave — and to hear about it from your sales team, which is a feature rather than a fault.
Below one order value
Effectively cash-on-delivery with extra steps. Legitimate for an account you have decided to wind down, and corrosive if used as a general policy — people stop respecting a control that fires on every transaction.
The most common mistake is not setting limits too low. It is setting them once, at onboarding, and never again — so a customer who has tripled their volume is permanently obstructed, and a customer who has halved theirs is permanently over-exposed. A limit is a judgement with a shelf life. Review the top twenty accounts twice a year and you will have done more than most.
What is and is not built
What AWRA OpsHub does today
- A credit limit recorded against the customer, and an exposure calculation that includes the opening balance, every open unpaid invoice, and the invoice being raised.
- Settled and cancelled invoices excluded from exposure, so headroom returns as soon as a customer pays.
- The invoice under edit excluded from its own exposure check, so editing does not double-count.
- A hold requirement returned with the limit, the exposure before and after, and the exact amount it is over by — so the conversation with the customer is about a number rather than a feeling.
- Evaluated on both the web and the API invoice paths, so a mobile or integrated invoice is checked the same way as one keyed at a desk.
- Receivables ageing and collections reporting for the part of the cycle that comes after this one.
What it does not do
- A limit of zero means no checking at all — not zero credit. An unset limit is unlimited credit, and it looks like an empty field rather than a decision.
- The trigger is the invoice, not the order. A quotation or sales order can be raised freely; the check happens when it becomes an invoice, which is later than some businesses would like.
- No ageing-based rule. A customer 120 days overdue but comfortably under their limit is not held. Lateness and size are separate concerns and only size is enforced.
- No automatic credit limit suggestion from trading history — every limit is a human judgement, typed in.
- No scheduled limit review or expiry, so nothing reminds you that a limit is three years old.
- No customer-facing visibility of their own limit or headroom.
The ageing gap is the one to cover with process. Size and lateness are genuinely different risks — the customer quietly at 90% of a sensible limit for two years is fine, and the one at 40% who has not paid since March is not. The limit catches the first and is blind to the second, so the ageing report remains a thing a person reads.
Deciding what happens when a hold fires
Best
A named person decides, and the decision is recorded
One or two people can release a hold, and every release gets a line saying who, why and what was collected in exchange. The control keeps its meaning because the exceptions are visible and countable.
Practical
Release on part-payment, as a standing rule
The hold clears when enough comes in to bring exposure back under the limit. Turns an argument into arithmetic, and gives the salesperson something concrete to ask the customer for instead of asking you for a favour.
Minimum
Hold, ship nothing, escalate the same day
Crude and it works, provided somebody actually looks at held invoices daily. A hold nobody reviews becomes a lost order rather than a collected debt.
Avoid
Routine verbal overrides at the counter
The fastest way to make a limit meaningless. Once overriding is normal and unrecorded, you have the administrative cost of a control with none of the protection, and no way to demonstrate afterwards that anybody was watching.
The awkward truth about credit limits
They work by causing friction with your best-liked customers at the least convenient moment, which is precisely why they are almost always set too high or not at all. The businesses that use them well are not the ones with the tightest limits — they are the ones where the limit is a recorded number rather than an opinion, and where a release leaves a trace. Everything else is theatre.
Our take
Record a limit on every account that carries a balance, because an unset limit is unlimited credit and it looks like an empty field rather than a decision. Set it at two to three months of normal trade for established accounts and one month for newer ones, then actually review the top twenty twice a year. Understand that the check fires at invoicing rather than at order, and that it measures size rather than lateness — so the ageing report stays a thing a person reads. And decide in advance who may release a hold and what gets recorded when they do, because an unrecorded override is indistinguishable from having no control at all.
Stop the debt before it becomes a collection
Credit limits checked against real exposure — opening balance, open invoices and the invoice being raised — on both the web and API paths, with the amount over the limit stated plainly. Ageing and collections handle what comes after.
See plans & pricingFrequently asked questions
What does the credit check actually compare?
Total exposure after the new invoice, against the recorded limit. Exposure is the opening balance on the customer record, plus every open invoice still owing — excluding anything paid or cancelled — plus the invoice being raised. It returns the limit, the exposure before and after, and the exact amount it is over by, so the conversation with the customer is about a figure rather than a judgement.
What happens if we do not set a credit limit?
The customer has unlimited credit. A limit of zero means no checking at all, not zero credit — the evaluation returns immediately and the invoice proceeds at any size. That default is deliberate, since the alternative is that nobody can invoice on day one, but it means the risky account is the one with no limit recorded rather than the one with a low limit.
Does paying an invoice free up credit immediately?
Yes. Settled and cancelled invoices are excluded from the exposure calculation, so headroom returns as soon as a payment is recorded rather than at month end. This is what keeps the control usable on your best accounts — a customer who pays promptly should never notice their limit exists.
Is a customer who is months overdue automatically blocked?
No, and this is the important gap. The check measures size, not lateness. A customer 120 days overdue but comfortably under their limit will not be held, while one at 95% of their limit who always pays on time will be. Lateness stays a matter for the receivables ageing report and a person who reads it.
When in the sales cycle does the check happen?
At invoicing, on both the web and the API paths. Quotations and sales orders can be raised freely — the check fires when the transaction becomes an invoice. For businesses that commit stock or start work at order stage, that is later than ideal, so the order-stage judgement remains human.
How should we handle overrides?
Name one or two people who may release a hold and record why every time, ideally with what was collected in exchange. The strongest standing rule is release-on-part-payment: the hold clears when enough comes in to bring exposure under the limit, which turns a favour into arithmetic and gives the salesperson something specific to ask the customer for. Routine unrecorded verbal overrides give you the cost of a control with none of the protection.
