Normalised earnings are a range, not a number

The single earnings figure on a teaser is not a fact, it is a proposal. Rebuild it properly and what comes out is a band, not a number, because three or four of the inputs are judgements rather than measurements. Anyone quoting you one clean figure has either not done the work or has done it and picked the end of the range they liked.

Why a single number is the wrong output

A small business does not have an earnings figure sitting in it waiting to be discovered. It has a set of filed accounts prepared to minimise tax, an owner who has been paying for some personal things through the company, a salary line that may or may not reflect what the job is actually worth, and a few years of results that differ from each other for reasons nobody wrote down.

Turning that into a number a buyer can underwrite means making a series of calls, and each call has a defensible range rather than a right answer. Add four of those together and the honest output is a band. Presenting the midpoint as a fact does not remove the uncertainty. It just hides it somewhere it cannot be argued with.

This is not a criticism of sellers. Filed accounts are prepared for the tax office and they do that job correctly. The gap between what those accounts show and what a new owner would actually earn is normal, expected, and the entire reason the exercise exists. What matters is whether the gap gets closed carefully or optimistically.

What comes out before the number means anything

The bridge from declared profit to something a buyer can use runs in a fixed order. Each step is a judgement, and each one moves the answer.

1. A real salary for whoever runs it

   The biggest adjustment in almost every small deal, and the one covered in detail below. If the owner is paying themselves nothing, or three times a market rate for tax reasons, the declared profit is a work of fiction in one direction or the other.

2. Personal costs that leave with the owner

   The car, the phone, the family member on the payroll who does not work there, the holiday booked as a trade show. These add back, but only the ones you can evidence. An owner who cannot show you the invoice is asking you to pay a multiple on a story.

3. One-off items, in both directions

   The exceptional gain gets removed and everyone remembers to do that. The exceptional cost that was actually a deferred routine cost gets removed too, and that one is a trap: three years without a van replacement is not a saving, it is a bill that has not arrived.

4. Maintenance capital expenditure, put back

   Whatever the business must spend each year simply to keep operating at its current level. Ageing equipment and an investment backlog show up here, and this is where a flattering margin usually comes from.

5. Working capital, sized properly

   Not an adjustment to earnings but a cost of ownership, and it is routinely forgotten. A business that finances its customers for sixty days needs cash to do that, and if it arrives underfunded you supply it on day one.

Run those five and you rarely get a single answer, because at least three of them are ranges. What you get is a plausible low case and a plausible high case, and the spread between them is often wider than the difference between a good price and a bad one.

The imputed owner salary, done properly

Every guide mentions adding back an owner's salary. Almost none of them say what number to use, which is where the whole calculation quietly goes wrong. The German trades sector has an actual published answer, and it is more rigorous than anything I have seen written in English.

The AWH standard is the valuation handbook used across the German crafts sector, recognised by the chambers, the banks and tax advisors. It sets out how to build the imputed owner's salary rather than leaving it to taste, and the construction is worth stealing wherever you operate.

Three things about that construction are worth noticing. It starts from a published wage rather than an opinion, so the floor is not negotiable. It requires a written justification for where in the 20 to 50 percent band you landed, which forces the judgement into the open instead of burying it. And the whole thing is deducted for family members working in the business too, at what a non-family employee doing the same job would cost.

If a business only produces a profit because the person running it is paid less than the job is worth, then it does not produce a profit. It produces a wage, and you are being asked to buy a job.

That is the point of the exercise and it is why the salary comes first in the bridge. A great many small businesses that look modestly profitable are, once a proper salary goes in, breaking even. That is not a disaster and it does not always kill a deal, but it changes completely what you are buying and what you can pay.

Then the standard cuts the number again

Normalising earnings is only the first half. The same handbook then sets out how to build the rate those earnings get capitalised at, and this is where the most striking figure in the whole document appears.

The capitalisation rate has four parts, and setting them beside each other is the fastest way to see what the standard actually believes.

Look at the last two rows together. Every other risk in the model, all eight of them, at their maximum, comes to twenty-four points. Owner dependency alone goes to thirty. The standard is saying that how much a business leans on one person can outweigh the entire rest of the risk assessment combined, and it is saying it in a published handbook rather than as a negotiating position.

The ten criteria it grades are worth reading as a list, because they are also a fairly exact description of what a buyer is checking on a first visit:

• The important customers, and whose relationship they are
• The main suppliers, and who holds the terms
• The bank relationships
• Technical know-how
• Commercial know-how
• Product and range decisions
• Workflow control
• Whether the owner works productively in the business rather than on it
• Whether any deputy or second management layer exists at all
• Their personal standing in the local community

Eight of those ten are transferable with enough time and a handover. Two are not quickly transferable at any price: technical know-how held in one head, and standing in a community built over thirty years. Which of the ten a business fails on therefore matters as much as how many, and that distinction is not in the score.

Sit with that for a moment. The professional standard used by German chambers of crafts says that how much a business depends on its owner can add thirty points to the rate at which its earnings are discounted, against a base rate of a little over two. Everything else in the model, the entire market and sector risk assessment, fits in a range a fraction of that size.

This is the strongest evidence I know of for something usually argued qualitatively. Owner dependence is billed twice, and the second bill is not a vague discount applied by a nervous buyer. In the German trades it is a documented, quantified, ten-criterion assessment that can quietly halve what a business is worth, and it is applied by the valuer the seller hired.

What to do with the band

Once you have a range rather than a number, the temptation is to average it and move on. Do not. The range is the useful output and it should stay a range right through to the offer.

• Underwrite the low case, and check that the deal still works there before anything else
• Take a haircut on top for the parts you could not verify, and size it by how many items were unevidenced rather than by feel
• Never let the high case set the price, because the high case is a forecast about your own future performance
• Write down which assumption each end of the band depends on, so you know what to test in diligence
• Check the spread itself: an unusually wide band is a finding about concentration or owner dependence, not a measurement problem
• Size working capital separately and add it to the price you are actually paying, because it is cash you will provide either way

The second and third points are where discipline is genuinely hard. A seller will always argue the high case, and the argument is usually reasonable, because the high case is what the business does when everything goes right and everything did go right last year. The question is not whether the high case is achievable. It is who should be paid for achieving it, and the answer is whoever does the work, which is you, afterwards.

The honest limit of all of this

None of the above produces a correct valuation, and treating it as though it does is a different kind of error from the one it is meant to prevent. What it produces is a defensible range with the judgements written down, which is enough to decide whether to proceed and how much to risk. Precision beyond that is false, and false precision is what makes people overpay confidently.

It also does nothing about the thing that most often ruins these deals, which is that the earnings were real but not durable. A rebuilt figure tells you what the business made. It does not tell you whether the customer that produced a third of it is renewing, or whether the only qualified technician is retiring. Those are separate questions, they are binary rather than numerical, and they belong before the valuation work rather than after it. If you are on the other side of this and want the same bridge run on your own numbers before anyone else runs it for you, that is worth doing early.

Valuation mechanics from the Handbuch Unternehmensbewertung für Handwerk und Mittelstand, AWH-Standard, published by the German chambers of crafts, which references IDW S1 for the risk-premium method. Base rates are the Deutsche Bundesbank yield-curve series the standard specifies. Not valuation or tax advice.