General contractors · Germany and Spain

General contracting, 2026

On a 7% margin, small leaks take a large share of the profit.

7.0%

That was the average return on sales in German construction in 2024, down from 10.2% in 2020. A general contractor with €15M turnover earns about €1M a year. Field rework alone averages 0.76% of contract value: about €114,000. On some projects it reaches 7.3%, the whole margin.

The company’s own books show about a quarter of it. Roughly half of rework starts with poor communication or project data: a decision that did not reach the site the way it was made.

One of my clients found out about a large loss on a single project only after handover. Every decision on it had been made correctly. Nobody could see that one of them had not been carried out.

Sources: DSGV analysis of about 10,000 balance sheets, published by Hauptverband der Deutschen Bauindustrie, March 2026 · P. Love, Journal of Construction Engineering and Management, 2026 · FMI and PlanGrid survey of 599 construction leaders, 2018.

What CoManager does

It keeps every decision on a project open until the site confirms it.

A general contractor makes hundreds of decisions per project between the office, the site manager and subcontractors: change orders, supplier swaps, spec changes, payment holds. Each one rests on an assumption about the site that nobody checks again. CoManager writes the decision down once, names what it depends on, and asks the person who owns it one short question at the right moment.

  1. 01

    Record. The decision, the alternatives rejected, and the assumptions it rests on. For example: “Supplier B delivers plumbing fixtures for floors 3 to 6 by 12 May at the agreed price.”

  2. 02

    Name the owner. One person commits to each assumption. A person, not the project.

  3. 03

    Ask, do not collect. On the review date the owner gets one question: “Delivered as agreed?” Answered from the site, not from a report.

  4. 04

    Separate two confirmations. “Confirmed by report” and “confirmed by outcome” are different states. Most losses hide between them.

  5. 05

    Raise the flag. When an assumption fails or nobody can answer, the owner of the company sees it the same week, not at handover.

  6. 06

    Seal. Each closed decision is dated and sealed. Over a year the company builds a record of which assumptions hold and which keep failing.

The one rule. CoManager never decides and never polices people. It shows where a decision and the site have diverged. A human decides what to do.

See it on one example company

An owner’s morning: six decisions, a phone for each person on the team, and cash by week with the gaps in red. Every button works; figures are fictional.

Open the demo

One decision, two ways

The same decision changes its outcome when someone writes down what it should achieve and checks.

Made up, but typical. Every general contractor has had all three.

The client changes the spec at a site meeting.

Usually

The site manager agrees on the spot. The extra cost is “to be settled later”. At handover the client disputes it, and the company absorbs it.

With a record and a question

The change is recorded with one assumption: “the client confirms the extra price in writing before work starts”. A week later the owner is asked: “Confirmed?” The answer is no, so the work waits for the signature.

A subcontractor sends an invoice for the month.

Usually

The quantities match the estimate, the name is familiar, the invoice is paid on a phone call. Months later the reconciliation shows part of it was never built.

With a record and a question

The rule is one line: “no invoice without the site manager’s measurement”. An invoice without it shows up in the morning, before payment, not at reconciliation.

A long-standing client offers a project at a thin margin, and cash is tight.

Usually

The owner says yes on Friday evening, tired. Half a year later the margin is gone, and nobody remembers what the decision rested on.

With a record and a question

The decision is recorded with what it rests on: “materials stay within 3%, the crew holds the schedule”. Each assumption gets a review date. When one fails, the owner sees it in month two, while there is still room to renegotiate.

A rule you can use tomorrow. Before any decision above a sum you choose, write one line: what result you expect, by when, and what must be true for it. On that date, check. If a decision does not change the result you expect, do not spend a meeting on it.

Who it is for

Owner led general contractors with 30 to 150 people and €7M to €30M turnover.

You will recognise this

  • Several projects run at once and you cannot be on every site.
  • There are two layers between you and the work: project managers and subcontractors.
  • You learn about a cost overrun when it is already a fact.
  • You judge progress from cameras, calls and weekly reports, and you know those three disagree.

Not a fit

  • Under 30 people: the owner already sees everything.
  • Groups where buying a new tool takes six months of procurement.
  • Companies looking for a new ERP or a cheaper task tracker.

What it is not

It sits on top of what you already run.

Looks likeThe difference
Procore, PlanRadarThey store drawings, defects, documents and tasks. CoManager stores decisions: why, on what assumptions, who committed, what came of it. Nothing is replaced.
ERPERP records what has happened. CoManager tracks what was agreed would happen, and whether it did.
Task trackerA task is already a consequence. CoManager works one step earlier, before a decision becomes a task.
ConsultingA consultant leaves after ten days. The system stays in the company and keeps asking the questions.
“AI that decides”It does not. It raises a flag. Your people decide.

Where AI fits

AI does not know the right decision for your company. It can gather the context.

Many owners expect a chatbot to tell them what to do. A model that has never seen your projects, your crews and your clients can only give an average answer from the internet.

What it can do from day one is collect the context: what was decided, on what figures, what was promised, what happened. Once a few months of your own decisions are recorded and tied to their outcomes, it can say which option has held up in your company and which has not. Even then it suggests. You decide.

That is why setup takes three months. The first weeks set up the records and the questions. Then decisions accumulate with their outcomes. Only after that is there enough of your own history for any suggestion to be worth reading.

How we start

Three steps. Each one can be the last.

Step 1 · 10 working days

Founder Sprint

€4,500 + IVA

A decision loss audit of one live project. I reconstruct its critical decisions, find the ones nobody confirmed and put a number on what they cost. You get a written record and a choice: stop here, or build.

Step 2 · 4 to 6 weeks

Build

€6,000 + IVA

CoManager set up on that project: decision records, owners, review dates, flags to you. Fixed scope; anything outside it is priced separately.

Step 3 · 3 months

Commissioning

€1,500 a month + IVA

Three months of tuning on your projects: your decision classes, your sources, your team trained. After that the same €1,500 a month keeps the loop running, or you stop.

Fixed price means fixed scope. Half at the start, half on delivery. No equity, no success fees.

Twenty minutes

Bring one change that cost you money. I will tell you whether this would have caught it.

Five questions: what changed, when you found out, where the original decision was written, who was responsible for passing it on, what it cost. If the answers are clear, you do not need the audit. If they are not, that is the audit.

Pavel Mihutski. Twenty years in operations and IT, 130+ projects in nine industries, and a construction materials company I founded at twenty. Based in Valencia, working in Germany and Spain. More about me

Two or three sentences are enough: what changed and when you found out.4000 characters left

Or write directly: Telegram · WhatsApp · LinkedIn