Notes from the floor.

Practical engineering notes from all four divisions — written for sourcing managers and plant engineers, not for search engines.

GANI ManufacturingGANI Contract ManufacturingGANI EngineeringGANI Lifecycle

2026-08-11 · GANI MANUFACTURING

What a golden sample actually freezes

A golden sample is not a nice part in a drawer. It is a contract about tolerances, process settings and acceptance criteria — and it only works if everyone agrees on what it freezes.

Read the article

2026-07-23

Retrofit without a shutdown: how migration windows are planned

Controls migrations fail on planning, not on wiring. How we structure pre-builds, parallel installations and fallback points so a PLC retrofit fits inside a weekend instead of a shutdown.

2026-06-30

The RFQ package that gets you a real quote in days

The difference between a quote in three days and a quote in three weeks is usually the buyer’s package, not the supplier’s workload. Here is what we need to price with confidence.

2026-06-09

Capacity extension: buying output, not machines

When demand outgrows a plant, the reflex is to buy equipment. Often the faster and cheaper answer is to buy guaranteed output from a partner — if the interface is defined properly.

2026-05-19

Why your 20-a-day variant doesn’t belong on your 1000-a-day line

Exotic variants cost far more than their share of line time. The changeovers, the error risk and the material handling all argue for splitting them out — and the numbers usually agree.

2026-04-28

What a FAT actually proves — and what only the SAT can

A factory acceptance test with a real script removes most commissioning risk before shipment. But some things can only be proven on site — and pretending otherwise inflates both tests.

2026-04-07

Kitting: where assembly errors go to die

Moving picking errors out of the assembly cell and into a controlled kitting process is one of the cheapest quality improvements available. Most plants still leave it on the table.

2026-03-17

The three spares lists every plant should keep

One undifferentiated spares list produces both overstock and eight-week line-down events. Splitting spares into three categories with different rules fixes most of it.

2026-02-24

100% testing or sampling: an honest decision tree

Sampling answers a different question than 100% testing, and the wrong choice either wastes money or ships defects with statistical blessing. The decision follows from failure cost, not habit.

2026-02-03

One source of truth: why we build panels from the EPLAN model, not from PDFs

When the schematic, the terminal plan and the BOM are outputs of one model, they cannot contradict each other. When they are separate documents, they eventually will.

2026-01-13

IEC 61439 and UL 508A: designing one panel for two markets

A panel built for the EU cannot simply ship to a US plant with different stickers. The two standards diverge on components, ratings and verification — and the divergence is manageable if you design for it from the start.

2025-12-16

SAT punch lists that actually close

Every site acceptance test ends with a punch list. The difference between a two-week closure and a six-month standoff is decided by how the list is written on the last day of the SAT.

2025-11-25

Serial traceability: what it costs, and what it pays back

Unit-level traceability sounds like bureaucracy until the first field issue. Then it is the difference between recalling forty units and recalling four thousand.

2025-11-04

Obsolescence is a calendar problem, not a crisis

Control components announce their death years in advance. Plants that treat those announcements as calendar entries retrofit on their own terms; plants that ignore them retrofit during a breakdown.

2025-10-14

Last-time-buy without the guesswork

A last-time-buy is a one-shot forecast with no correction possible. Sizing it from consumption data, failure history and remaining machine life beats sizing it from fear.

2025-09-23

Maintenance windows that hold: planning service around production

A maintenance window that overruns teaches a plant to stop granting windows. Scoping, pre-staging and a hard fallback rule are what keep the window credible.

2026-08-18

How to transfer a running product without anyone noticing

A transfer is a risk decision, not a price decision. The measure of success is not the unit cost at the new supplier — it is whether your customers can tell that anything happened.

2026-08-14

The stock bridge: the least glamorous part of a transfer

Nobody puts the overlap quantity in a presentation. It is still the number that decides whether your transfer is remembered as a project or as an incident.

2026-08-08

Gap analysis: what the old supplier never wrote down

The drawings are only the part of the process that someone bothered to record. The gap analysis is the search for everything else.

2026-08-04

First article approval: what you are actually signing

A first article report is not paperwork for the file. It is the moment where a drawing becomes a manufacturing agreement — and it deserves to be read that way.

2026-07-30

How a recurring panel family is released

The first panel of a family is an engineering exercise. Panels two through four hundred are a documentation exercise — and that is the part that decides your cost.

2026-07-24

Revision control when the build happens somewhere else

In-house, a revision change is a conversation. Outsourced, it is a transaction — and treating it as a conversation is how mixed-revision stock happens.

2026-07-18

Deviation requests, done honestly

Every factory deviates sometimes. The difference between a good supplier and a bad one is whether you hear about it before the units ship or after.

2026-07-12

Approving an alternative component without surprises

An alternate that matches the datasheet can still fail in your machine. The approval process exists to find out which differences matter.

2026-07-06

How a test fixture earns trust

A fixture that passes bad units is worse than no fixture at all, because it converts a quality problem into a documented false confidence.

2026-06-30

The arithmetic of a line stop versus a buffer

Buffer stock looks like waste on a balance sheet and like oxygen on a shop floor. The argument is usually won by whoever brings the arithmetic.

2026-06-24

Material liability: the clause nobody reads until it hurts

Long-lead material has to be bought before the order that consumes it. Who carries that exposure — and for how long — belongs in writing.

2026-06-18

What a manufacturing frame agreement should actually say

Price and payment terms take one page. The clauses that decide whether the relationship works take the other four.

2026-06-12

Directed buy: who owns the risk when you pick the part

Directing a component is a legitimate engineering decision. It also moves a piece of the supply risk back onto your side of the table, and that should be explicit.

2026-06-06

Incoming inspection is not a formality

The cheapest place to find a wrong part is the goods-in bench. The most expensive place is your customer's plant.

2026-05-30

Reading a supplier quality questionnaire honestly

Most questionnaires can be answered to look excellent. The useful ones are answered to be accurate — and a good buyer can tell the difference in ten minutes.

2026-05-24

Building a second source without splitting your engineering

A second source protects supply. Badly implemented, it doubles your engineering load and halves the attention each supplier gets.

2026-05-18

Box build: what changes when you hand over the whole product

Outsourcing a subassembly is a capacity decision. Outsourcing the finished product is an organizational one, and it changes what your own team does all day.

2026-05-12

Firmware and serial numbers: the pair that has to stay together

A serialized product with unversioned firmware is only half traceable — and it is always the wrong half when a field problem appears.

2026-05-06

Conformal coating: when it earns its cost

Coating is not a universal upgrade. It is a targeted answer to a specific environment, and applying it without that environment in mind buys cost rather than reliability.

2026-04-30

Harness documentation that survives five years

Harnesses are the assemblies most often built from tribal knowledge — and the ones most likely to be re-ordered long after the person who knew has left.

2026-04-24

Crimp quality: what is worth measuring

A crimp is a gas-tight metal joint made in a tenth of a second. Everything you will ever know about it comes from a handful of checks — chosen well or chosen badly.

2026-04-18

The pilot batch is not a prototype

A prototype proves the product can exist. A pilot batch proves the process can repeat it — and confusing the two is how ramps go wrong.

2026-04-12

Ramp plans that do not break the supplier

A ramp that looks smooth on a chart can be brutal on a floor. The realistic curve is the one built from the pilot, not from the demand plan.

2026-04-06

Why NRE, tooling and unit price belong on separate lines

A single all-in price hides where the money goes and makes every later change an argument. Splitting the quotation is a small honesty with large consequences.