Feasibility first: ten questions before a vision PoC

Why “can you inspect this?” is not a brief

A photograph of a part and a wish for “100% inspection” is a conversation starter, not a project. Vision succeeds when someone owns the decision the system must make every cycle — and accepts what “good enough” means under real takt.

The ten questions we ask

1. What decision is required every cycle? Pass/fail, classify, measure, locate for a robot, read a code — pick one primary job.
2. What do good and bad look like? Bring real OK, NOK and borderline samples. Golden parts alone lie.
3. Cycle time and miss budget? A 200 ms vision job with a 2% miss allowance is a different system than a 40 ms job with zero silent misses.
4. Geometry of the station? Working distance, FOV, motion, vibration, access for a light and cable.
5. Lighting reality? Glare, washdown, ambient light, space for a dome or backlight — lighting is often the project.
6. Which modality is mandatory? Presence, print, codes, OCR, 3D, AI — do not stack all of them “because we might need it”.
7. Interfaces? PLC brand, robot brand, MES hooks, discrete I/O versus industrial Ethernet.
8. Who owns the recipe after go-live? Operator changeovers, format changes, who may retrain a model.
9. Environment? Temperature, ESD, oil, hygiene class, enclosure rating.
10. What makes the PoC a clear pass or fail? Write the success criteria before you buy hardware.

Walk these with production and controls in the same room. Use our feasibility checklist as the shared worksheet.

What good preparation looks like

Photos of the station from operator height. Reject samples in a labelled bag. A sketch of the handshake. A named owner for changeovers. That package turns a demo into engineering.

Engagement after the brief

Depending on depth, Sedeco works full service, as line integration partner, or as value-added component supplier — see How we work.