Industry Information

When a Vision Inspection Machine Is Not Worth Buying (Yet)

Four situations where a vision inspection machine is not worth buying yet - volume sampling covers, defects cameras cannot see, parts outside the bands, and no defect history to configure against.

hzxx-jc01-single-table-standard-model

The Question the Brochure Cannot Answer For You

Most of the selection guides on this site start from a purchase already in motion - this one covers the other exit: the situations where a vision inspection machine is not worth buying yet, and what to do instead. A supplier telling you this sounds counterintuitive, but the reasoning is practical: a machine installed before its preconditions exist sits underused, gets configured against guesses, and becomes the reason "we tried automation and it did not work". If none of the four situations below applies to you, the selection guides are the next stop. If one does, it is cheaper to know now.

Situation 1: Your Volume and Risk Are Already Covered by Sampling

100% machine inspection earns its cost (/posts/vision-inspection-machine-cost-7-price-factors) where the risk is concentrated - the risk-concentration cases the sampling article (/posts/sampling-plans-vs-full-inspection) lists. Where the risk is thin and the volume is low, an AQL sampling plan already covers the exposure, and a station rated 120-840 pcs/min per the brochure would sit idle between batches. The honest test is your claim history, because claims are the only signal that shows what sampling missed: if the last two years of customer claims show no pattern your sampling plan failed to catch, the machine has nothing to catch that your plan is not already catching.

Situation 2: The Defects That Matter Are Outside Camera Scope

Cameras check what cameras can see. The established boundaries on this site: fully internal voids, same-color material mixing, and material-level properties are not surface-appearance questions; a crack that stays closed while the part is relaxed may need compression-style checking - the brochure's compression machine (HZXX-JC04, the extrusion-line unit) checks 8-35 mm rings under applied pressure precisely for this class. If the defects driving your purchase decision are mostly in that invisible set, an appearance inspection machine will pass the parts your customer complains about - which is the worst possible first experience. Write your defect list first (how to define OK/NG criteria (/posts/defining-defects-ai-visual-inspection)), and check honestly how much of it is surface-visible before you shop for cameras.

Situation 3: Your Parts Do Not Fit the Bands

Each machine has a brochure size band: 6-80 mm or 6-120 mm OD on the turntable machines, 8-35 mm on the compression machine, 100-400 mm and 300-1000 mm on the tensile lines, thickness limits per model - among others, and the by-size guide to O-ring inspection machines (/posts/o-ring-inspection-machine-by-size) maps the rest. If your parts sit outside every band, the answer is customization, not a standard machine - and customization means engineering time and a configuration that has no prior proof. Also treat part presentation as a fit question: a part that will not rest and present its faces predictably on a turntable needs a configuration answer proven at a sample test, not assumed from the catalog.

Situation 4: You Have No Defect History to Configure Against

The machine is configured to your quality standard - which presupposes you have one in writing. With no NG log, no claim history and no defined defect families, the configuration meeting has nothing to configure to: thresholds become the supplier's guesses, and the acceptance test proves nothing about your actual risk. The same problem shows up at the sample test: a useful sample batch is built from your worst-case defective parts and borderline pieces - if you cannot produce those, you cannot validate detection either. Build the defect specification first; it costs sorting discipline and quality-team hours, not capital - and it is the input every later step consumes.

What to Do Instead

• Let sampling carry the current risk while you build three months of sorted NG data - the record that later justifies or kills the purchase.
• Fix process-level causes first: inspection screens parts, it does not change the mold, the material or the press settings. The machine's own NG statistics are a monitoring tool, not a process fix.
• Run the sample test anyway if you are unsure - it replaces assumptions with evidence before any capital decision: your parts, the images and the OK/NG decision log returned to you, per the sample test walkthrough (/posts/vision-inspection-sample-test). A failed or unimpressive sample test is a fast, evidence-backed no - far cheaper than discovering the same answer after installation.
• Revisit when a trigger changes the picture: a new customer audit, a claim pattern - which flips the question from daily-volume justification to a 100% containment re-screen, covered in re-screening a rejected lot (/posts/re-screening-rejected-lot-containment) - or volume growth that moves the sampling-vs-100% balance.

Frequently Asked Questions

Is manual inspection cheaper for small volumes?

Often, yes - that comparison (labor cost structure versus machine cost structure, consistency, records) is covered in manual vs vision inspection (/posts/manual-vs-vision-inspection). The point of this article is the step before it: whether automation is warranted at all yet.

What volume justifies a vision inspection machine?

There is no honest universal number - it depends on your rates, part value, claim exposure and the machine class in question. Anyone quoting you a universal threshold without your numbers is guessing.

My NG logs look clean - doesn't that prove sampling is working?

A clean NG log shows your sampling caught what it tested for - it cannot show what it missed. Escapes surface as customer claims; that is the containment scenario covered in re-screening a rejected lot (/posts/re-screening-rejected-lot-containment).

Will the machine reduce my defect rate?

No - it screens parts and produces records. The defect rate comes down when the process cause is found and fixed; the machine's NG statistics point at which family is growing.

Can I buy now and configure later?

Configuration to guessed thresholds is how machines end up distrustful fixtures. The proven sequence is the one the sample test implements: defect list first, then configuration, then evidence.

Next Step

Sort your last NG logs by defect family, check how much of your risk is surface-visible, and measure your parts against the bands - if none of the four situations above fits your picture, the selection guides are your next stop; if one does, you have just saved yourself a purchase that would not have worked. Contact the team (/contact-us) - Monica (+86 138 5726 5705, WhatsApp) or Arya (+86 173 2430 9735), or email monica@aivisioncheck.com.

Read next

More from the inspection knowledge base

More machine vision guides and product spotlights from our engineering team.