Good company, bad ops
Business quality and fixable operations are two separate axes, not one score. The money is in the corner where a good company is run badly.
The business you want is a genuinely good company that is being run badly. Not a cheap company, not a broken one you plan to rescue. Those are two different things and collapsing them into a single score is the most expensive mistake in small acquisitions, because a single score lets a weak business with obvious problems outrank a strong business with quiet ones.
One score hides the thing you are looking for
Rank targets on one number and something strange happens. A mediocre company with a terrible website, no CRM and no marketing scores well, because on a checklist of fixable problems it has the most fixable problems. Meanwhile a well-run company with a fifteen-year maintenance book and a slightly dated website scores lower, because there is less obviously wrong with it. The ranking is upside down, and it stays upside down no matter how carefully you tune the weights.
The fix is not better weighting. It is recognising that you are measuring two unrelated things and forcing them through one number. How good the business is has almost nothing to do with how badly it is currently run. Those are independent, and once you separate them the map you actually need appears.
The two axes
Quality is how good the company is regardless of who runs it: recurring work, earnings you can verify, customers who are not going anywhere, people who can do the job. Operating upside is how much money is currently trapped by doing everything by hand. High on both is the target. High quality and low upside is fully priced. Low quality and high upside is the trap.
What you want is the corner where both are high, because that is where the gap is widest between what the seller can prove and what the business produces once it is run properly. The seller can only be paid for what the business does today. Everything the manual back office is currently losing is real, present, and invisible in the accounts, which means it is not in the price.
Three failures that look identical from outside
The whole judgement rests on telling these apart, and the words get used loosely, so here they are pinned down.
- A bad business
- The market does not want enough of what it sells at a price that leaves a margin. No amount of operating skill fixes this, because the problem is demand rather than execution. Buying one is buying a job with debt attached.
- A good business, badly run
- Customers want it and pay for it, and the money leaks between the enquiry and the invoice. The demand is proven, which is the expensive half to create, and what is missing is the cheap half to install.
- A tired business
- Both were true once and the owner stopped. Real demand, decayed operations, and nobody intending to rebuild. This is the one most often mistaken for the first, and it is usually the second.
Only the middle one is worth paying for, and the third is where the bargains are, because it gets priced as the first.
Quality is scarcer than the market suggests
The instinctive objection to a strict quality bar is that it leaves nothing to buy. In a market where a demographic wave is pushing owners toward the exit, surely there is no shortage. The data says the opposite, and it says it more sharply than most people in this field seem to realise.
Germany's Institut für Mittelstandsforschung publishes the standard estimate of how many businesses are approaching a handover. In its 2025 edition, around 186,000 companies are expected to change hands between 2026 and 2030. That is the number everybody quotes. The number almost nobody quotes sits two pages later: of roughly 3.2 million family businesses in Germany, only about 733,000 earn enough profit to be classified as worth taking over at all. Under a quarter.
Then compare it with the previous edition. For the 2022 to 2026 window the same institute counted 772,000 businesses worth taking over out of 3.3 million. So in one estimation cycle the pool of buyable companies shrank by roughly 39,000, and that shrinkage is the stated reason the succession count did not rise even though owners kept getting older. Demographics pushed the number up. Deteriorating earnings pushed it down harder.
The wave everyone is waiting for is not a wave of good businesses. It is a wave of businesses, of which about one in four is worth owning.
That single fact reorganises the whole exercise. If three quarters of what comes to market fails a basic earnings test, then deal flow is not the constraint and never was. Selection is. Any process that optimises for seeing more targets rather than for rejecting faster is solving a problem you do not have.
What actually goes on the quality axis
Quality has to be built from signals a seller cannot manufacture in the twelve months before a sale, and cannot fix quickly even if they wanted to. That rules out most of what appears on a teaser. Revenue growth can be bought. Margin can be flattered by deferring maintenance. A tidy website means nothing at all.
How much of the book is contracted or recurring
The single heaviest signal, because it is the one that determines whether next year exists before anyone sells anything. Maintenance agreements, inspection obligations and framework contracts, as a share of revenue. Below a quarter is a project business wearing a service business costume.
Whether the earnings survive being rebuilt
Take the declared figure apart and put it back together: a market salary for whoever actually runs it, real working capital, owner expenses removed, one-off gains stripped. The gap between the teaser number and the rebuilt number tells you more about the seller than any conversation will.
How durable the customers are, past the concentration test
Concentration above a quarter is a separate binary check that already ended the conversation. This is the question underneath it: how long has the average relationship lasted, is there a contract, and does the customer have a reason to stay that is not personal loyalty to the departing owner.
How deep the qualified bench is
Again, past the binary gate. One additional licensed person means the deal can complete. Three means the business can absorb somebody leaving in year two, which is a completely different risk.
Whether the core is genuinely hard to displace
Physical work, on site, requiring certification, with equipment that has to be inspected on a cycle. This is what makes the improvement you are about to make permanent instead of temporary.
Margin belongs on this axis too, but with one important adjustment that people get wrong in both directions. A thin margin caused by structurally low prices in a commodity market is a quality problem and should be scored as one. A thin margin caused by three people doing by hand what a system should do is not a quality problem at all. It is the opportunity, and it belongs on the other axis entirely. Penalising it twice, once as weak margin and once as the thing you are there to fix, is how a good target gets ranked below a bad one.
What goes on the operating axis, and the mistake in measuring it
The second axis is where almost everyone builds it wrong, myself included on the first attempt. The obvious approach is to score whether the operational gaps exist: no CRM, no marketing, no automated invoicing, tick, tick, tick. Do that across a realistic pool of offline companies and every single one scores near the top, because having a manual back office is close to universal in this population. A signal that fires on everything is not a signal. It carries no information and it separates nothing.
What matters is not whether the gap exists but how much money is falling through it, which depends almost entirely on the size of the business underneath. A five-person operation missing half its calls is losing a small amount of money. A forty-person operation missing a quarter of them is losing a great deal. Same tick on the checklist, entirely different prize.
The second adjustment is that this axis has to be weighted lower than quality, and deliberately so. Operating upside is a forecast about what you will manage to do after completion. Quality is an observation about what the business already does. When those two disagree, the observation should win, because forecasts about your own future competence are the least reliable input in the entire process and the most flattering to believe.
The trap in the other corner
Everything above exists to keep one specific business out: the weak company with enormous obvious upside. It is genuinely seductive. The problems are visible, the fixes are ones you know how to make, and the price is low precisely because everyone else can see the problems too.
The reason it fails is that a business with poor quality has poor quality for a reason, and the reason is almost never the back office. It is thin structural margins, or customers who are there on price and will leave on price, or a trade anybody can enter with a van and a phone. Fix the operations of a business like that and you have an efficiently run company in a bad position. The efficiency is real. It does not change the position.
Which is why the quality axis functions as a floor rather than a contributor. A target does not become worth pursuing by scoring spectacularly on operating upside. It has to clear a minimum on quality first, and only then does the upside score decide where it sits in the queue. In practice that turns a long list into a short one very quickly, and the short list is the only thing worth spending diligence money on. The binary checks that run before any of this are in the seven things that end a deal on the first call, and what happens to the operating half after completion is in the product cannot be disrupted, the operations can.
How to run it without building anything
None of this needs software. It needs a sheet with two columns of scores, a composite that keeps quality senior, and the discipline to record every target you looked at including the ones you rejected. The rejected rows are the valuable ones, because after forty of them you can score a new target against how comparable businesses actually turned out rather than against your opinion in the moment.
One honest limitation, since this is where these systems oversell themselves. You cannot calibrate the weights on outcomes. Buying a handful of companies over several years produces far too few results to fit anything to, and any claim that a scoring model learns from closed deals at this volume is arithmetic that does not work. What you can calibrate on is the plentiful data: how targets looked at screening versus what diligence found. That is a periodic human review of your own guesses, written into a file, not a model that improves on its own. Anyone describing it the other way is selling the software. If you want the version of this that a seller sees, it is what I actually look for when I buy.
The short version
- Business quality and operating upside are independent. Forcing them into one score ranks weak companies with obvious problems above strong companies with quiet ones.
- Of roughly 3.2 million German family businesses, about 733,000 earn enough to be classed as worth taking over. The pool shrank by around 39,000 in one estimation cycle, which is why the succession count stopped rising.
- Deal flow is not the constraint. Selection is.
- Score the magnitude of what the manual back office is losing, not the fact that it exists. Every offline company of this size has gaps, so their presence separates nothing.
- Quality is a floor, not a contributor. A high operating score never promotes a business that failed the quality bar.
Questions I get on this
How many businesses are actually worth buying?
Is a business with no website and no CRM a good acquisition target?
Why not buy a cheap struggling business and turn it around?
Succession and business-count figures from the Institut für Mittelstandsforschung Bonn, Daten und Fakten Nr. 37, "Unternehmensnachfolgen in Deutschland 2026 bis 2030" (2025), compared against Nr. 27 for 2022 to 2026.