Introduction

Close the leaks.
See the margin.

MarginOS is job-level margin accounting for project work: it shows which jobs ended up unprofitable, where the difference came from, and which rows prove it — using the systems you already have.

demo.marginos.net · questions: jarmo@thecloudonline.com

What this presentation covers

The service and how it is used, end to end

Move with ← → , the space bar, or a swipe. The contents list jumps straight to a slide.

01 · What MarginOS is

Job-level margin that survives an audit

You quote a job on certain assumptions, and the work departs from them. MarginOS computes the margin on every job, breaks the gap between quoted and actual down to the euro — and shows, for every figure, the row it was computed from.

The promise

We find the leaks and show what they rest on. The arithmetic is deterministic and traceable to source data, and no figure is shown without a row a reader can check.

02 · Who it is for

Small and mid-sized companies doing project work

Manufacturing

  • Subcontract machine shops
  • Contract manufacturing

An example: demo.marginos.net

Technical services and contracting

  • Maintenance and property services
  • Electrical and HVAC contracting
  • Construction and installation

Expert work

  • Project offices
  • Consulting and design services

The precondition is not the age of your system but whether jobs, hours, material and invoicing can be exported in some form. After machine shops, the rules are fitted together with you — that is part of the first conversation.

03 · The problem

The margin is lost inside the job, not in the pricing

The estimate is made to sell

A quote is built from an hours estimate and material prices that are the best guesses available at the time.

The actual accumulates over months

Hours, material, subcontracting and rework are posted into different systems at different times.

The difference surfaces too late

By the time it shows, the job has shipped and nothing can be done about it. The same mistake goes into the next quote.

The problem is not a missing calculation. It is that the estimate and the actual live in different places and are never compared, job by job.

04 · How it works

From source data to findings

1

Import

Jobs, time entries, material consumption, subcontracting, other variable costs and invoicing come in from your existing systems.

2

Compute

Every job gets its quoted margin, its actual margin and the difference, under the agreed rules.

3

Raise

The rules raise the jobs whose difference passes the agreed threshold — in euros, not as an opinion.

4

Show

The queue says where to look first. Every finding opens onto its source rows.

05 · The exceptions queue

Where to look first

Ordered by money

Severity first, then the euro amount. The top row is where the most money is at stake — not whatever happened to catch the eye.

Every row states its reason

A finding is not an alert but a sentence: which rule fired, how much money is involved, and which rows demonstrate it.

Filterable

The queue can be narrowed by rule or work centre — and a filtered view is a link you can send to a colleague.

No action without a margin

A job whose margin cannot be computed raises no monetary findings at all. It is shown as blocked, not guessed at.

06 · The margin bridge

To the euro, where the difference came from

On a job's own page the quoted and the actual are broken into the drivers that account for the whole gap:

The drivers

  • Revenue difference
  • Material quantity and material price
  • Labour hours and labour rate
  • Subcontracting and other variable costs
  • Rework

An identity that holds

The drivers sum to exactly the difference between quoted and actual. Whatever the data does not explain is shown as its own driver — unexplained, not zero.

The bridge's figures are in the demo: demo.marginos.net

07 · Evidence

Every figure opens onto its source rows

A finding is not an assertion but a conclusion you can take apart. Under every figure are the rows it was computed from.

A row states

  • Which table and which system
  • Which identity and version
  • Which period
  • Under which allocation rule
  • In which quality state

Two roles, kept apart

Some rows produced the amount; others only show that the job exists. They are listed separately, so a reader knows which row actually moved the figure.

08 · A blocked job

When the data is not enough, no figure is shown

If a job is completed and invoiced but its material consumption or its hours were never posted, there is no margin — only a guess. So no margin is printed.

What a blocked job shows

  • A dash where the margin would be — not a zero, not an estimate
  • The reason recorded the same way every other finding is
  • Evidence that opens onto the source rows

Why this is a feature

A figure nobody can check is worse than a missing one: it reaches a report and gets used to decide something. A blocked job also says which posting is missing — it is a to-do list, not an error message.

09 · Accounting rules

The rules are agreed before the first figure

What is agreed

  • Which jobs are in scope
  • How subcontracting and other variable costs are allocated
  • Which thresholds raise a finding
  • Which exceptions block a margin from being shown

What is not negotiable

  • The arithmetic is deterministic: the same data and the same rules give the same figure
  • Every figure is traceable to a source row
  • The rule version and its validity are visible in the interface

Your finance side approves the rules before the first figure is shown.

10 · The data

What is needed — and what is not

Needed

  • Jobs and what was quoted on them
  • Time entries
  • Material consumption
  • Subcontracting and other variable costs
  • Invoices and credits

Not needed

  • Replacing your systems
  • New bookkeeping
  • Write access to your database

The source data remains your property. The import is read-only, and the arithmetic runs on the data you have handed over.

11 · Getting started

From the first job to a measured result

1
A first conversation. We walk through one completed job and see whether the difference would have been found. Thirty minutes, no commitment.
2
Mapping the data. We find where jobs, hours, material and invoicing can be exported and in what shape.
3
Approving the rules. We fit the rules to your industry and your practice, and your finance side approves them.
4
Running the history. We compute an agreed set of completed jobs and go through the findings together.
5
Measuring. We agree what gets measured, so the benefit does not stay an opinion.

12 · Technology

Arithmetic you can check

A deterministic core

Money is never computed in floating point. Amounts are integers in cents and hours in hundredths, so rounding does not accumulate and the same data cannot give two different answers.

Traceability in the structure

Every imported row carries its source system, identity and version. That is not an appendix to the report; it is a precondition of the arithmetic.

No black box

The rules can be read in the interface, and every figure opens onto the row it was computed from.

13 · Where to begin

Three ways forward

1
Look at the demo: demo.marginos.net — 120 completed jobs with the leaks still in them. The data is synthetic, not a customer result.
2
Book a first conversation: jarmo@thecloudonline.com — thirty minutes, no commitment.
3
We are looking for the first two or three companies to define the rules with, run their history through, and measure the result.

The first version is read-only: it shows the difference and opens every figure onto its source rows. The approval workflow and the verification of an action come next, together with the first customers.

Try it

See it for yourself.

demo.marginos.net — an open demo, no sign-in.

Questions: jarmo@thecloudonline.com

Suomeksi