Legacy Scale Panel Reading

The problem

The scale with the smallest quantities carries the biggest risk.

The scale that weighs minor components (additives, colorants, flavors in grams) has been on the floor for over a decade with a proprietary display that talks to nothing, and carries the highest risk of formula deviation.

• Colorants, flavors, vitamins, enzymes and processing aids are weighed in grams on a scale that has been on the floor for over a decade. Its display is proprietary and it has no port, network or export the plant can use.
• These are the ingredients where one extra gram is a formula deviation: an overdosed colorant, a vitamin claim the batch no longer meets, an additive above its legal limit in the customer's finished product.
• The operator reads the display and writes the figure on paper. If the reading is rounded, swapped with the next ingredient or skipped, nothing records it and nobody finds out until the lab assay, if it is assayed at all.
• Replacing the scale means capex, revalidation and recalibration downtime on a line that works, so the decision keeps getting postponed while the risk stays exactly where it is.

How it fits the IRIS system

Connect in photo mode on the panel — read from outside, the scale untouched.

Connect photo mode on the panel: an external camera reads every weighing with OCR and a grounded model, matching it against the active formula sheet, without touching the original panel.

The scale is not opened, rewired or recalibrated. A camera on a bracket reads what the operator already reads, and the difference is that every reading is kept, compared with the active formula and attached to the batch. The replacement decision stops being urgent. And because the reading is matched with the formula as it happens, a wrong addition is caught while the batch is still in the weighing booth, not when the lab assay comes back out of range.

See the full IRIS architecture →

Before and after

The minor-ingredient scale today versus the scale being read

Aspect	Today	With iLEAN Connect
Gram-level record of each addition	Handwritten, if written	Every stable reading captured
Match against the active formula sheet	Done by eye	Automatic, ingredient by ingredient
Overdosed colorant or vitamin	Found at the lab or by the customer	Flagged at the scale
Scale calibration and validation	—	Untouched: nothing installed on it
Minor-ingredient scale replacement capex	Pending in the plan	Deferrable indefinitely
Traceability of additives and flavors	Gaps by the gram	Complete, gram by gram

Minor-component weighing with no digital record → full gram-by-gram batch traceability. Scale replacement indefinitely deferrable.

Impact estimate

Impact estimate — to be validated with your numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 5-10 months, mostly from avoided recalibration and replacement capex on the minor-ingredient scale.
• Fewer minor-component deviations, which are the ones that turn a compliant premix into a non-conforming one with a single misread gram.
• Full gram-by-gram batch traceability for the additives, colorants and flavors that today leave no digital trace at all.
• And the scale replacement moves from the urgent column to whenever it actually makes sense for the plant.

Payback 5-10 months · avoids recalibration CAPEX and cuts minor-component deviations — estimate to validate

And the fair question from the production manager

"Can a camera really read a gram-level weight off an old display?" — it is an anchored task: a fixed display, digits or a needle in a known position and a value that must fall inside the tolerance of the ingredient being weighed, where the best models drop below 1.5% error [1]. A frame that is blurred, glared or outside the possible range is discarded and read again rather than accepted, and a persistent doubt is sent to the operator.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.