One customer at 40% of revenue is a cost-of-capital problem

If one customer is 40% of your revenue, the answer is not a threshold. Concentration does not reduce the cash your business produces. It raises the rate at which everyone else discounts that cash — your bank, an investor, a buyer. That is how the multiple falls, and it is charging you already, years before any sale.

It is also three separate risks wearing one name, and most owners measure only the first of them.

Why the usual answer is the wrong shape

Search this question and you will find bands. One mid-market M&A advisory publishes a set: above 20% of revenue from a single customer is elevated risk; 20–30% costs half a turn to a full turn of EBITDA; 30–50% costs one to two turns; above 50% means restructuring the business or walking away from the sale.

The bands are plausible and they may well be right. They are also, by the author’s own explicit statement, drawn from personal deal experience with no data behind them. I have no objection to experience. It is frequently the best thing in the room. I object to experience wearing the clothes of evidence, because a range stated firmly enough gets repeated until it sounds measured.

What the evidence actually establishes

There is peer-reviewed work here. Dhaliwal, Judd, Serfling and Shaikh, writing in the Journal of Accounting and Economics in 2016, examined the relationship between customer concentration and the cost of capital. They found a positive association between customer concentration and the cost of equity, and a positive association between customer concentration and the cost of debt. The effect was stronger among suppliers more likely to lose a major customer. It survived propensity-score matching and instrumental-variable estimation, which is the authors’ answer to the obvious objection — that risky firms might simply end up with concentrated customers rather than the other way round.

I am going to state the direction and stop there. I am not going to quote you a magnitude in basis points, because that figure is not one I have verified at source, and an invented number would be worth less to you than no number at all.

One further limit is worth naming. Studies of the cost of equity are run on companies whose cost of equity can be observed, which means listed ones. Your company is not listed. What transfers to you is the mechanism, not the coefficient.

Why a discount rate is not an abstraction

Here is the arithmetic that makes the mechanism concrete. A stream of cash flows growing at a steady rate is worth, roughly, one year of cash flow divided by the difference between the discount rate and the growth rate. Hold growth at 2% and move the rate.

Illustrative only: what a percentage point of discount rate does to a multiple, at 2% growth. Not the study’s findings — the shape of the mechanism.

Discount rate	Growth rate	Implied multiple of cash flow
12%	2%	10.0×
13%	2%	9.1×
14%	2%	8.3×

One point on the rate takes roughly 9% off the value. Two points take about 17%. I have chosen one and two points because they are convenient numbers to divide by, not because they are anybody’s finding about concentration. The point of the table is that nothing happened to the business inside it. Same revenue, same profit, same customers, same everything. Only the risk assessment moved, and a tenth of the value went with it.

Concentration does not reduce the money your company makes. It raises the price everyone else charges you for the possibility that it stops.

It is already priced into your bank facility

The cost-of-debt half of that finding is the part with immediate consequences, because you do not have to be selling anything for it to apply. Your credit margin, the size of facility you are offered, the covenants attached to it and the collateral asked for are all outputs of the same judgement, made by somebody with a spreadsheet who can see your debtor ledger.

Which produces a quiet asymmetry. Your bank has priced your concentration. You have not. They will not open the conversation, because from where they sit nothing is wrong: the risk has been identified and it is being compensated.

Concentration is three risks, not one

Revenue concentration is the one everybody measures — the share of turnover that leaves when they leave. It is the least interesting of the three.

Margin concentration is the share of profit that leaves. Those are different numbers and often startlingly different, because the largest account is usually the one that was discounted hardest to win and has been discounted since to keep. I have seen companies where the dominant customer contributed a far smaller share of contribution than of revenue, and cases where the contribution was close to nothing once cost to serve was allocated honestly. That changes the nature of the problem completely: you are not defending a profit stream, you are defending a capacity filler. Working out which of the two you have requires margin by customer, which almost no SME has in front of it.

Decision concentration is the one nobody measures and the one that does the lasting damage. It is the point at which a customer sets your specifications, your terms, your production calendar, your investment priorities and eventually your strategy. It arrives gradually and every individual step is reasonable. By the time an account is setting your production calendar it is not really a customer. It is a division of your company that you do not control, cannot price and cannot sell.

A turn of EBITDA is a nameable amount of money

The Argos Index, which tracks eurozone mid-market transactions, put the first quarter of 2026 at 8.6× EV/EBITDA, up from 8.3× in the final quarter of 2025. Within the same quarter, private equity buyers paid 10.0× and strategic buyers 7.8×.

2.2 turns separated private equity buyers (10.0×) from strategic buyers (7.8×) in eurozone mid-market deals in the first quarter of 2026 — different buyers, same market, same quarter.

That index covers deals of €15–500m, which is to say companies substantially larger than a typical Swiss SME, and its absolute level does not transfer to you. What transfers is the unit. A turn is one year of EBITDA. On a company earning CHF 1,200,000 of EBITDA, a turn is CHF 1,200,000 and half a turn is CHF 600,000, which is more than most owners have spent on outside advice in the entire life of the business. Which multiples actually apply to a company your size is a longer and less comfortable subject.

So how much is too much?

There is no evidence-backed threshold, and I would rather say so than invent one. What does have an answer, inside your own data, are three questions that matter considerably more than the percentage.

The first is how substitutable you are to them. Not how important they are to you — that is the number owners compute, and it runs in the wrong direction. If your part is specified into their product, tooled to their line and qualified through their audit, a 45% dependence may be far more durable than a 25% dependence on a customer who could change supplier by email.

The second is what the account contributes after everything you give it. A large customer at a real contribution margin is an asset. The same customer at a contribution margin near zero is a capacity constraint you are paying to keep.

The third is how long the notice period is, in the contract and in reality. Twelve months of notice turns an existential risk into a difficult year. Ninety days does not.

What this means in practice

Diversification is not a decision. It is eighteen to thirty-six months of deliberate commercial work, and the timescale is not consultant’s padding. It follows from what has to happen.

A second channel takes that long because it has to be built while the first one still absorbs your entire commercial capacity. Repricing the base takes that long because the customers currently subsidising your largest account have to absorb an increase in sequence, with the ones most likely to leave tested early rather than last. And a pipeline takes that long because the customers capable of replacing a large account are, by definition, large themselves, and large customers have procurement cycles measured in years.

All three depend on doing the work while the concentrated account is still paying for the lights. That is the uncomfortable part. The moment when diversification becomes obviously urgent is the moment it stops being affordable.

What this cannot tell you from the outside

Whether your 40% is a serious problem or a manageable one depends on facts I cannot see from here: the contribution margin of that account after cost to serve, the switching cost in both directions, the notice period, how many of your other customers are priced as though that account did not exist, and whether the relationship belongs to the company or to one person who is sixty-four.

Notice who is not producing those facts. Your bank has priced the risk and will not be sharing its workings. Your accountant records the revenue accurately and has not been asked to split contribution by customer. Your lawyer read the notice period without being asked what it implies for your cost of capital, because that was not the instruction. Every one of them is doing their job correctly.

What a single customer is genuinely worth to you, and what their departure would genuinely cost, falls between all of those jobs. Which is a fair description of the work nobody has been asked to do.