Insights
Notes on method
2026-09-05 · 5 min readWhere spreadsheet scheduling runs out: when constraint-based scheduling is worth itA spreadsheet is a good scheduling tool right up until the constraints start touching each other. The signals that decide when production scheduling software is worth it — and when the spreadsheet should be left alone.→2026-09-05 · 4 min readWhich job actually pays? The data layer that shows margin at quote timeIn most companies profitability is learned at month end, as a single lump total. What cost items and what definition layer it takes for per-product profitability to work at the moment a quote is made.→2026-09-05 · 5 min readWhat OEE is, how it is calculated, and why one number is not enoughOEE multiplies three ratios: availability, performance and quality. The product is a single number, and on its own it does not say why it fell. Without the components and coded stop reasons, the metric does not drive a decision.→2026-09-05 · 5 min readWhere does the scrap go? Marking production loss with data instead of instinctIn most workshops scrap is known but not counted. Before reducing loss you have to mark it where it happens: a definition, a measurement point, and a record kept in the moment.→2026-08-20 · 1 min readWhen can you trust an AI's answer? The accuracy gate and source citationAn AI model always produces an answer; the real question is when that answer should be shown to the user. Two simple mechanisms: a threshold and a source.→2026-08-20 · 1 min readRebuild the ERP from scratch, or build on what you have? Decision criteria“Let's scrap it all and start over” is a tempting but expensive sentence. A few concrete questions decide when to keep the current system and when to replace it.→2026-08-20 · 1 min readWhat “a good model” means in demand forecasting — and how it's measuredWhether a forecast model works is understood not from the number it prints, but from how it is measured. How to test accuracy against past data, not on paper.→