What I look at before I look at the numbers

Before a single figure changes hands I can tell you most of what I need to know about a small service business from outside it, in about twenty minutes, using nothing the owner has to send me. The read is not a substitute for diligence. It decides whether diligence is worth paying for, which on a list of two thousand companies is the decision that actually matters.

Why the outside read comes first

Accounts arrive late in the process, after a conversation, sometimes after a signed non-disclosure. That is fine for one deal and impossible for a pipeline. If the only way to assess a target is to get its numbers, then your capacity is however many owners will send you numbers, which is a small number and a slow one.

So the first pass has to run on what is publicly observable. The useful thing about a small trade business is that a great deal is observable: how it presents itself, how customers reach it, what customers say afterwards, whether anything automated sits behind the front door, and whether the company is hiring. None of that is the business. All of it is evidence about how the business is run.

It is also the only part of the process that scales without permission, which matters more than it sounds. Every other step depends on somebody agreeing to talk to you. This one does not, which means it can run continuously in the background against a whole sector while you are busy with something else.

The finding nobody expects

Here is where most people build this wrong, and where I built it wrong first. The intuition is that an old-fashioned business will be obvious because it has no website. That intuition is out of date by roughly a decade, and acting on it will make you reject exactly the companies you should be looking at.

Bitkom and the German trades association surveyed 504 trade businesses in 2025. Ninety-four percent have their own website. Eighty-five percent offer at least one digital service. Sixty-two percent send invoices digitally. Fifty-six percent use cloud tools. On the surface, this sector is not behind at all.

That split is the whole read. You are not looking for a company that failed to get online. You are looking for a company that got online in 2014, stopped there, and has been running the actual work by phone and paper ever since. Those are extremely common, they look perfectly respectable from the outside, and they are invisible to anyone screening for the absence of a website.

The website is a poster. What matters is whether anything is standing behind it, and in most of these companies nothing is.

The eight things I actually check

Each of these takes two or three minutes and produces a note rather than a score. The point is to build a picture of a real business with an untouched operating layer, which is a specific combination rather than a list of complaints.

1. What the site can actually do

   Not how it looks. Can you book anything, get a price, or start a job without picking up the phone. In almost every case the answer is no, and the contact route is a phone number and a generic mailbox. That is a brochure, not a front door.

2. Reviews, and specifically their rhythm

   The rating matters less than the pattern. Consistently strong reviews arriving at a slow, irregular trickle means the work is good and nobody has ever systematically asked. That is the ideal shape. A cluster of reviews all posted in the same fortnight two years ago means somebody ran a campaign once and stopped.

3. Whether the phone gets answered

   Call during working hours, from a normal number, and see what happens. Voicemail on a Tuesday at eleven is a finding. So is a ring-out with no voicemail at all, which means the enquiry left no trace anywhere.

4. What the technology footprint says

   What the site is built on, when it was last substantially changed, whether there is any booking, chat, tracking or CRM behind it. A site untouched since 2016 with no analytics on it has never been treated as a channel.

5. Whether anyone is buying attention

   Ad transparency registers are public in both search and social. Almost none of these companies appear. That is not a criticism, it is an indication that every customer they have arrived by referral, repeat work, or the phone book, which tells you something good about the underlying business.

6. The registry and the filings

   Legal form, who the managing directors are, how long they have been there, and whether accounts have been filed on time. Late or missing filings are a flag about how the back office is run long before anyone opens the numbers.

7. Job postings, read backwards

   An active listing for a qualified technician tells you the bench is thin and they know it. No listings at all, over years, tells you either that they are stable or that they have never hired through a channel you can see. Both are worth knowing, neither is decisive.

8. What the equipment implies

   What they install and service determines whether there is a legal inspection or maintenance cycle attached to it. That cycle is what turns a project business into a book of recurring work, and you can often infer it from the brands and services listed on the site.

Read together, the pattern you are hoping to find is boring and specific: a company with strong sparse reviews, a decade-old brochure site, no advertising, clean filings, and equipment that has to be inspected annually. Everything in that list says the business is real and the demand takes care of itself. Nothing in it says anybody has ever built a system.

The list is easier to trust once you see it run. Below are the two shapes that come up most often. Both are good businesses, which is the point: the difference is not quality, it is whether the upside is still there.

The row that does the most work there is the sixth. A holding company appearing in the ownership chain is the cheapest available signal that somebody professional arrived before you did, and it is public. Everything downstream of that point, the ads, the booking tool, the tracking, is the consequence rather than the cause. Finding it early saves an hour that was never going to go anywhere.

The inversion, and why it matters

There is one rule here that most screening processes get exactly backwards, and it is worth stating on its own because it changes what your data means.

When you look up one of these companies and find nothing, no analytics, no CRM, no ad history, no job postings, the natural instinct is to record it as missing data. Your sheet has a column, the column is empty, and empty columns feel like incomplete research. Then, because the target has so little information attached to it, it drifts down the list in favour of a company you could find more about.

So the fields have to be treated as two different kinds. A blank on something backward-looking and verifiable, filed accounts, customer concentration, whether the maintenance book is real, is a genuine hole in your knowledge, and the target should be marked as low confidence until it is filled. A blank on a forward-looking operational field, no CRM, no ads, no software, is not a hole at all. It is a positive result and should be recorded as one.

Verifiable blank

A fact about the past that exists somewhere and you have not obtained yet. Filed accounts, contract terms, concentration. Score it low and flag the uncertainty, because you genuinely do not know.

On-thesis blank

An absence in the operating layer that is itself the evidence: no booking system, no CRM, no advertising, no automation. Nothing is missing from your research. The thing you were looking for is the fact that it is not there.

Getting this distinction wrong is not a small scoring error. It systematically inverts the ranking, pushing the best targets to the bottom of the list and promoting the ones that have already had their upside taken. Any screening sheet with a single "data completeness" column is making this mistake by construction.

What the read cannot tell you

Being honest about the limits is what keeps this useful. The outside read is silent on almost everything that determines whether a deal is possible. It cannot see customer concentration, and concentration is the most common reason a good-looking business is unbuyable. It cannot see whether the earnings survive being rebuilt. It cannot tell you whether the owner is the only qualified person, or what is buried under the yard, or whether the seller wants a handover or an auction.

All of those are in the binary checks that end a deal on the first call, and every one of them needs a human conversation. The outside read exists to earn that conversation, nothing more. Treating it as a verdict is how you end up confidently pursuing a company that fails on the first question you ask out loud.

It is also worth saying plainly what is not acceptable here. In Germany the boundaries are real: you cannot harvest a listing board wholesale, and unsolicited automated commercial contact to businesses is not a grey area. The read is for prioritising who is worth approaching and how, and the approach itself stays a letter and a person. Anyone selling you a scraper that ends in an automated outreach sequence is selling you a legal problem with a dashboard on it.

Where this fits

The sequence is narrow on purpose. The outside read takes two thousand companies to perhaps two hundred worth a look, the binary checks take that to a few dozen, and the two-axis score decides which of those get a conversation. Diligence money only ever lands after all three, on a target that is already pre-qualified as a good business with an untouched operating layer.

The reason to run it in that order is cost. The read is nearly free and can run against a whole sector. Diligence is expensive and can only run against one company at a time. Any process that spends the expensive resource before the free one is inverted, and it will be slower and worse than a process that does two thousand cheap reads first.

There is a catch in that sentence worth doing the arithmetic on, because it is the reason this looks easy and is not.

So the read is only free at the scale where it matters if most of it is not done by a person. Seven of the eight checks are collection: what the site runs on, what the registers hold, when the filings landed, whether an ad account exists, what the reviews look like over time. That is retrieval and it should be machine work, arriving as a filled sheet. The eighth, the phone call, is not, and neither is the judgement at the end, which is a human looking at nine assembled facts and deciding whether they add up to a real business with an untouched engine room.

That division is the whole reason the process works: the machine does not decide anything, it removes the reason not to look. The twenty minutes then get spent on the twenty companies that earned them instead of being spread one at a time across two thousand that mostly did not.

The scoring layer that sits on top is in good company, bad ops, and if you are on the other side of this and wondering how a buyer sees your business before you have said a word, that is roughly what I am looking at.

Digitalisation figures from Bitkom Research and the ZDH, Digitalisierung des Handwerks (2025, n=504, representative sample of German trade businesses).