The cheapest production line is the one you don't have to rework after three months of running. Cheapness isn't in the price, but in how many meta-decisions were made correctly in the early stage — when they could still be fixed in CAD, not in the hall.
You'll find these six phases in every project that goes to production. Some projects skip one — and then pay.
Phase 1 — Analysis of the process, not of machines
The client usually starts with "which injection molder?" This is the wrong question. The right question: which products do you want to make, in which series, with which materials, at what cycle time, what dimensions? The injection molder then chooses itself.
Without this, you stand in front of 30 catalogs and 30 prices — no criterion for decision. This is exactly where an independent partner is worth the most. A vendor tied to a brand will recommend their brand. An independent one will recommend what actually fits.
Phase 2 — Layout before technology
Make a paper layout. Sometimes literally paper — at real scale, on the hall floor. Mark material entries, finished product exits, operator passes, maintenance access, evacuation access.
Only then do you look at whether specific machines fit into it. If they don't, the layout changes — not the technology. If the layout is bad, the machines become enemies of operators and maintenance.
Phase 3 — Media (services) as the primary investment
Air, water, gas, electricity, exhaust, data. The cost of services is usually 15–25% of total investment and the hardest to change after installation. If air is routed twice wrong, the second fix costs ten times the first, because there's a machine in between and production around it.
Invest 20% more in services than seems necessary. The reserve will get used within 18 months.
Phase 4 — Installation in an order that doesn't get swapped
The heaviest machine, the largest equipment — first. First mechanics, then electrical installation, then automation, then integration. This order doesn't get swapped, because the last phase needs the previous three as a stable foundation.
If you try to parallel-track electrical and mechanical, you save 1–2 weeks. You lose them during commissioning, because someone damaged a freshly laid cable while moving a machine.
Phase 5 — FAT before SAT, always
Factory Acceptance Test at the machine supplier. Site Acceptance Test at your hall. If you skip FAT, you do SAT with an unverified machine — and every fault is debugged on your floor, during your hours, at your cost.
FAT takes 2–5 days and costs travel and hotel for your engineer. SAT after a skipped FAT takes an extra week and costs that week's missing production.
Phase 6 — Ramp-up with metric transparency
The first month of production is not production. It is tuning. Plan it as tuning, not as production. Measure OEE from day one — not to have a number, but to see which of the Six Big Losses is keeping the line under 90%.
Most common causes in the first month: small stops (unexpected stops < 5 min), startup losses (slow ramp after every shift change), reduced speed (the line runs at the speed of the slowest machine). Each of these losses is solved by a different measure — and all three are easier to solve than material defects (which are rare in a stabilized process).
What of this is paid?
In practice: the first three phases are consulting — they're usually included in the design fee, but if you skip them, the client pays twice. Phases 4 and 5 are "execution" — most clearly measurable. Phase 6 is borderline — sometimes part of the contract, sometimes separate.
If you're choosing a partner, ask which of these phases they guarantee and which they only recommend. The machine vendor guarantees the machine — they have no responsibility for the layout of your hall. The integrator guarantees the process — but only if it's in the written scope.
---
*We write this from our own practice and from real post-mortem meetings with clients who had one of the phases skipped. If you're interested in a concrete project, we'll walk these phases applied to your numbers on a 30-min call.*