I look at how your process runs and rebuild what you order: production code that runs in your systems.

Forward Deployed Engineer · Salzburg, Austria

One day on site. I walk one order from intake to invoice and note where the time goes. At the end you get a results briefing: what can change, in which order, and at what price. What gets built is your decision afterwards.

Request a walkthrough — €4,000 fixed

The survey

One order flow across three sheets. The figures here are a worked example and were measured at no client; they show which quantities a process survey yields: duration per leg, waiting time, rework rate.

01

Current state (survey)

Sheet 01 — Current state: the process as measured A node diagram of today’s order flow. Intake passes through three systems (ERP, production planning, CRM) and two manual re-entries into an approval queue, and from there to invoicing and close. A dashed exception loop runs from the approval queue back to the start. The boxes sit unevenly and the edges cross three times. Every leg is dimensioned with its measured duration; the critical path sums to 96 hours, or 4 days. Order intake ERPCreate order Manual re-entry I MESProduction planning Manual re-entry II Approvalqueue CRMShipping order Invoiceand close 6 h 4 h 9 h 5 h 3 h 2 h 17 % rework 7 h 69 h 96 h lead time · 4 days 3 systems · 2 manual re-entries · 17 % rework
The process as it runs today: one intake, three systems, two manual re-entries, an approval queue and an exception loop back to the start.

Measured legs

  1. Order intake → ERP, create order — 6 h
  2. ERP → manual re-entry I — 4 h
  3. Manual re-entry I → MES, production planning — 9 h
  4. MES → manual re-entry II — 5 h
  5. Manual re-entry II → approval queue — 3 h
  6. Approval queue → invoice and close — 69 h (waiting)
  7. Branch: manual re-entry II → CRM, shipping order — 7 h
  8. CRM → invoice and close — 2 h
  9. Exception: approval back to the start of the process — 17 % (rework)

Critical path: 96 h, or 4 days. 3 systems, 2 manual re-entries.

96 hours, Lead time today — 4 days
02

Measurement

Sheet 02 — Measurement: the same process, marked up The same drawing as sheet 01, unchanged in geometry, with a red correction layer. The critical path is traced in a heavy stroke and sums to 96 hours. Of those, 69 hours are pure waiting time in the approval queue — 72 percent of the lead time. Three exception loops are circled freehand and labelled A1, A2 and A3: the two manual re-entries and the jump back to the start of the process. Nothing is solved on this sheet. Order intake ERPCreate order Manual re-entry I MESProduction planning Manual re-entry II Approvalqueue CRMShipping order Invoiceand close 6 h 4 h 9 h 5 h 3 h 2 h 17 % rework 7 h 69 h A1 A2 A3 69 h waiting 72 % of lead time Critical path 96 h · 69 h of it waiting 3 exception loops marked · nothing solved yet
The same drawing, nothing moved, marked up in red. The critical path is traced heavy, the waiting time is broken out, and the three exception loops are circled freehand and labelled A1 to A3.

Findings

  1. Critical path, longest route through the flow — 96 h
  2. A1 — exception loop at manual re-entry I — 17 % (rework)
  3. A2 — exception loop at manual re-entry II — 2 systems (entered twice)
  4. A3 — jump from approval back to the start — 1 loop
  5. Waiting time in the approval queue — 69 h (72 % of lead time)

Of 96 h lead time, 69 h is pure waiting.

69 hours, Waiting time — 72 % of lead time
03

Target state (rebuild)

Sheet 03 — Target state: the rebuilt process The same nodes, redrawn as a straight run with no crossings. Both manual re-entries are gone and the exception loop back to the start is closed. Two steps are now handled by a model; they are drawn dashed and blue and marked AI in the text: extracting the order into the ERP, and triaging exceptions before approval. The legs measure 0.5 hours, 0.5 hours, 1 hour, 3 hours and 1 hour. The dimension line at the foot sets the old lead time of 96 hours, or 4 days, greyed and struck through, against the new one of 6 hours. Order intake ERP · AIOrder created Manual re-entry I MESProduction planning Manual re-entry II Approval · AIExceptions triaged CRMShipping order Invoiceand close 0.5 h 0.5 h 1 h 3 h 1 h AI extracts the order AI triages the exceptions 96 h · 4 days 6 h lead time 3 systems · 0 manual re-entries · 0 exception loops
Nodes are merged, both manual re-entries are gone, and the exception loop is closed. Two steps are taken over by a model; they are drawn dashed and marked “AI” in the text. The other three steps stay with people.

Legs after the rebuild

  1. Order intake → ERP · AI, order created — 0.5 h (AI extracts the order)
  2. ERP → MES, production planning — 0.5 h
  3. MES → approval · AI, exceptions triaged — 1 h (AI triages the exceptions)
  4. Approval → CRM, shipping order — 3 h
  5. CRM → invoice and close — 1 h

Lead time: 6 h. 3 systems, 0 manual re-entries, 0 exception loops.

6 hours, Lead time after the rebuild

Lead time 4 days 6 hours

Drawing key

What that means

Points you can put to any provider. With me they hold for every engagement.

Write access, not a presentation
I work in your production systems. At the end I hand over a repository with a passing test suite.
Reachable at 7am
When it jams during the early shift you call me, not a support line. I wrote the code.
It keeps running at your place, without me
The code runs in your environment and sits in your repository, together with the evals and the runbook. Nothing switches off when the contract ends. Your team does the next rebuild, or whoever you hand it to. Which rights belong to whom is settled in the contract, not on a website.

Ask every provider you compare the same question: who writes the code that ends up running?

Engagements

What the first day costs is published here, because the day is a known quantity. For anything after it I quote neither a price nor a range in advance. Put a figure on work you have not seen and you are guessing, and the margin for guessing ends up on your invoice. The results briefing states, for every proposal, how many working days it takes and what it costs. All prices are net of Austrian VAT; for business clients outside Austria, VAT is reverse-charged.

Walkthrough

€4,000 fixed

One day at your site. I walk one order through, talk to the people who run it daily, and look at the systems it passes through.

Extent

1 day on site

What you hold afterwards

A briefing: the drawing of your process, the places where time is lost, and the proposals that follow — what can change, roughly what it is worth, and what it costs. Each proposal can be ordered on its own: a few days for the smallest, several weeks for a rebuild, months where it turns into an ongoing engagement.

Rebuild

fixed price per proposal

I build what you order out of the briefing. Every item has its price before it starts: if the build takes longer than expected, I pay for that, not you.

What you hold afterwards

Running code in your environment, evals, runbook and a handover your team can carry on from.

Operations

from €4,400 per month

On call for what I built, plus a capped allowance of days for continued work. No retainer for availability alone.

What you hold afterwards

For the systems I wrote, I am the contact. Every month you get an account of where the days went.

Evidence

Three cases out of thirteen years, anonymised: industry and country, no names. Each one also says which part of it was mine. The long form of each is a sheet under References.

Energy and climate tech, European scale-up From the chip to the end of the line — and then handed over
Device software on the computer inside the heater, drivers for third-party mining hardware, the cloud broker between them, the Kubernetes platform, a PKI for device identity, and the test bench at the end of the production line. Since 2024 I have been the lead engineer on this system. At the end of 2026 it goes over entirely to the client’s in-house team, following a roadmap that included onboarding an internal engineering lead. The mobile app I led and carried rather than wrote.
~270,000 lines, hand-written code, ~11,500 commits
Healthcare, community pharmacies in Austria Four point-of-sale systems behind one interface
Oracle SQL, partly reverse-engineered from the database schema. IBM DB2 over ODBC, debugged on site in the pharmacy. Plus a vendor REST interface and a vendor web service. One model sits on top of those four technologies, and all four run in production, so that stock, terms and order status do not get typed twice. The central backend of that platform is mine from the first line to the last.
~7 years, in production
Furniture manufacturing, Austrian solid-wood supplier Construction rules that come out as machine data
A configurator checks made-to-measure wardrobes against the real production range, prices them live and emits one machine project per carcass, plus the cut lists. Those rules previously lived in people’s heads and in spreadsheets.
12+ cabinet types, parametrically constructed

See all references

A client name goes on this page once I have the written release for it, and you rarely get one in this business: non-disclosure agreements usually forbid it. So what stands here is industry and country. Who it was I will tell you in conversation.

The figures in this section are rounded counts from the source archives. Those in the plan set above are an example, and are marked as one there.

Contact

Write and tell me which process costs you the most time, and when you could clear a day for it. You get a proposed date and the engagement sheet back.

Request a walkthrough – office@mumbler.gmbh