Setting Customer Credit Limits Without a Credit Department

If you deliver work before you are paid for it, you are a lender. That is not a metaphor and it is not a way of framing the point dramatically. You have advanced value against a promise to pay later, on terms you probably did not negotiate, to a counterparty you probably did not assess, with no security and no ability to call the loan.

The decision to do that is being made in your company today. It is being made by whoever accepted the order, and it is being made without a record, which means nobody can tell you afterwards what was known at the time. The purpose of a credit matrix is not to start making credit decisions. It is to write down the ones that are already happening, at the moment they happen, by somebody whose job it is.

The version that works at this scale is small. Three or four bands, a short information list against each, a named approver, and a trigger that forces a re-look. It fits on one page and it takes an afternoon to build. Companies that skip it are not saving the afternoon; they are deferring it to the quarter in which a customer stops paying and nobody can find out who agreed to the exposure.

No dollar figures appear in this piece

Deliberately, and it is worth saying why rather than leaving it to be noticed.

The band boundaries are the one part of this that cannot be written for you. A limit that is prudent for a company with a strong cash position and a wide customer base is reckless for a company with a drawn operating line and three customers, and the difference is not a matter of degree. Any figure printed here would be adopted by somebody it does not fit, and it would carry more authority than it earned simply by having been printed.

So the bands below are defined by consequence rather than by amount, and the owner converts consequence into a number once, in a single sitting, using the company’s own cash position. That conversion is the only piece of arithmetic in the exercise and it is the only piece that is genuinely specific to the business.

Defining the bands

Ask the owner one question per band, in this order, and write the answers down as dollar amounts against each.

Band	The question that sets its ceiling
Routine	What size of loss would be absorbed in the ordinary course, annoying but not requiring a decision?
Considered	What size of loss would require the owner to be told immediately and would change the quarter?
Significant	What size of loss would force a change in plan: a delayed hire, a deferred purchase, a draw on the facility?
Board level	What size of loss would put a covenant, a payroll or the facility itself at risk?

Most owners answer the first and last quickly and hesitate on the middle two. That hesitation is useful information, and the answer to give them is that the middle boundaries can be wrong by a wide margin without much cost, whereas the top one cannot be wrong at all.

The output is four numbers, dated and signed, that live at the top of the matrix. Revisit them when the company’s cash position materially changes, not annually out of habit.

The matrix

Band	Information gathered before credit is extended	Approves	Standing review	If the customer will not provide the information
Routine	Legal name confirmed against the public corporate registry, billing detail sheet complete, terms accepted in writing	Finance, on the file	Balance and ageing reviewed in the monthly close like any other account	Proceed. The information here is administrative rather than diagnostic.
Considered	The above, plus two trade references contacted and their answers recorded, plus how long the customer has been trading	Finance lead	Reviewed at each close against the limit, and on any move up the collections ladder	Extend at the routine ceiling only, and revisit after a period of clean payment history with you
Significant	The above, plus the customer’s own financial statements where they will provide them, plus a documented view of what the customer’s failure would do to your quarter	Finance lead and owner, jointly	Formal review on a set cadence, and immediately on any missed promised payment date	Do not extend at this band. Offer one of the structures below instead.
Board level	The above, plus a decision recorded in writing by the owner naming the exposure, the reason it is acceptable, and what would cause it to be withdrawn	Owner, in writing	Reviewed every close, by name, on the collections page of the reporting pack	Do not extend. This band is not a negotiation.

Three notes on reading it.

The routine band has almost no diagnostic content on purpose. Gathering references for a small first order costs more than the exposure and it teaches the sales side that the credit process is an obstacle, which is how the process gets routed around. Make the bottom band nearly frictionless so that the top band can have real friction and still be respected.

Trade references are worth more than they are usually given credit for, and only if you actually contact them. A reference list supplied and filed unread is theatre. Two telephone calls asking a simple factual question, how long they have supplied the customer and whether payments arrive when due, produce a better answer than most of what can be bought, because a supplier who has been paid late will usually say so plainly.

The last column is the one that gets deleted from these matrices. It should not be. A customer’s refusal to provide information is itself information, and if the matrix has no answer for a refusal then the answer defaults to whatever the person under commercial pressure decides in the moment.

When the answer is no, offer a structure rather than a refusal

A credit matrix that only produces yes and no will be overruled by the owner, correctly, on the first good customer it declines. Give it a middle path.

Structure	What it does	When it fits
Deposit on order	Reduces exposure to the balance rather than the whole contract	Project work with a defined start
Payment in advance for the first period	Converts a credit decision into a cash transaction until history exists	New customers of any size
Shorter milestone spacing	Bills more often against the same contract value, so exposure never accumulates	Progress billed work
A limit that grows on evidence	Starts at the routine ceiling and steps up after a stated number of clean payment cycles, automatically, without a fresh approval	The most useful structure on this list, and the least used
Security or a guarantee	Changes the nature of the exposure entirely	A legal instrument with legal consequences. Raise it with the company’s lawyer and do not draft it internally.

The growing limit deserves the attention. It converts an argument about whether a new customer is creditworthy, which nobody can settle, into a rule about what they have to do to become so, which settles itself. It also removes the most common failure of these matrices, which is that a customer approved at the routine band in year one is still sitting at the routine band in year four while buying five times as much, so the real exposure has moved and the file has not.

The triggers that force a re-look

Limits are set at onboarding and then left. The customer that hurts a company at this size is almost never a new one; it is one approved years ago, under different conditions, by someone who has left.

Trigger	Why it matters	What it forces
The customer reaches its limit and asks for more work	The most obvious trigger and the one most often waived under commercial pressure	Re-approval at the band the new total sits in, not the old one
A promised payment date is missed	Behaviour has changed before the ageing shows it	Move to the next band’s information requirement
The account reaches rung 4 on the collections ladder	An account manager is now involved in the money conversation	Limit frozen pending review
Ownership, name or banking details change	The counterparty may not be the entity you assessed	Full re-verification of the legal entity
A trade reference or a supplier tells you they are being paid late	The market usually knows before your ledger does	Immediate review, whatever the band
The customer’s own industry turns	Correlated risk, and it arrives across several accounts at once	Review the group, not the account
Twelve months elapsed with no review at the significant or board band	Staleness is a cause of loss in its own right	Scheduled review, on the calendar rather than on request

The row about a supplier telling you is the one worth building a habit around. The information exists, it circulates informally, and finance is usually not in the conversation where it circulates. Ask your account managers, once a quarter, whether they have heard anything about how any of your customers are paying other people. It costs one question in a meeting.

Concentration is a separate decision, and it is the owner’s

A per-customer limit controls what you lose if one customer fails. It does nothing about what happens if the customer that fails is the one representing a large part of the revenue, because that exposure is not really a credit exposure at all. It is a business model exposure and the receivable is only the visible part of it.

No proportion is stated here, and none should be, because the right one depends on the company’s fixed cost base, its facility headroom and how quickly it could replace the work. What can be given is the method for arriving at it.

Take your largest customer. Assume they stop paying in the current week and stop buying at the same moment. Run that through the cash model built in the 13-week cash forecast piece: zero the receipts from that customer, leave the cost base where it is for as long as it would realistically take to act, and read the closing bank line and the headroom line week by week. Then do the same for the second largest.

The output is not a percentage. It is a week number, and it is the week in which the company would have to do something it does not want to do. The owner looks at that week and decides whether it is close enough to be intolerable. If it is, the cap follows from that judgement, and the cap belongs to the owner, because it is a decision about what kind of company this is going to be rather than a finance calculation.

Once set, the cap has to be visible where the decision to accept more work is actually made. A concentration limit that lives in a finance document and not in the sales conversation is not a limit.

The credit file

One page per customer, and it exists so that a question asked in eighteen months has an answer.

• The band and the limit, with the date and the approver’s name.
• The information gathered, and what it said. Not “references checked” but what the references actually reported.
• Any structure applied: deposit, advance, growing limit and its step-up rule.
• Every review since, dated, with the outcome even where the outcome was no change.
• The legal entity verification, refreshed on any name or ownership change.

If a customer fails, this page is what tells you whether the loss was a bad decision or a bad outcome. Those are different things and the difference matters, because one of them requires a change to the matrix and the other requires nothing at all. Without the file, every loss looks like the first kind and the response is usually to tighten credit across the whole ledger, which costs revenue from customers who were never the problem.

Where the matrix sits alongside the rest of the control set, including who is permitted to raise a credit note against a balance that has already been billed, is covered in the financial controls engagement.

What we do not do

We do not buy a scoring product for a company with sixty customers and then treat the score as the decision, because the score is an input and somebody still has to own the answer. We do not set a limit the sales side has never seen, since an invisible limit is discovered only when it is breached. We do not draft guarantees or security documents, which are the lawyer’s work. And we do not approve an exposure verbally, because the entire value of this exercise is that in eighteen months there is a piece of paper with a date and a name on it.