Pick-to-Light Software for Your WMS or ERP
Poka-yoke pick-to-light for kitting bays: the correct bin lights, the other bins stay physically locked until a bin sensor confirms the right pick, and kits resolve from ERP work orders so no operator ever chooses the variant.
Schematic
Pick-to-light at the pick face
What the operator sees at the bin, and the sequence the lights walk them through.
A poka-yoke pick-to-light system where the error-proofing is the lock, not the lamp
MileSoft Pick-to-Light turns a kitting bay into a control function. Each bin carries an indicator, an interlock and a sensor: only the correct bin lights with the quantity shown, adjacent bins stay locked until the right pick is confirmed, and confirmation comes from a bin sensor rather than a button press, so the record reflects what was actually taken. A wrong pick is prevented, not flagged.
Kits resolve from ERP and MES work orders over an ISA-95 boundary. A kit barcode starts the sequence and the station loads that unit's pick list, so no operator ever chooses the variant and a BOM revision reaches the line without re-teaching a station. Every pick is logged with authenticated operator, bin, kit and timestamp for audit playback, and a deviation holds the station and records its reason.
In a published deployment at a Tier-1 automotive supplier kitting six close-cousin radar-module variants, operator pick errors fell from about 1.4% to zero wrong-part picks since rollout, and new-operator ramp-up from four to six weeks to five days. Zero wrong picks is not zero wrong kits: wrong quantities and wrongly stocked bins remain, and they are the next thing to error-proof.
Poka-yoke pick-to-light system
The light is not the poka-yoke. The interlock is.
A lamp tells an operator which bin to open. Telling is a warning, and warnings depend on attention.
Shingo separated the control function of an error-proofing device — the process cannot continue when something is wrong — from the warning function, where the operator is signalled and may carry on. Warning reliability belongs to the person receiving it, and it degrades under exactly the conditions that produce errors: fatigue, time pressure, an unfamiliar variant, the end of a shift. On a pick-to-light bay the lamp signals and the lock controls; they are Shingo's two categories sitting side by side on the same bin.
| Component | Function | If it is removed |
|---|---|---|
| Indicator | Warning — which bin, and how many | Correct picks get slower; wrong picks stay impossible |
| Interlock | Control — the other bins do not open | Wrong picks become possible again; the guarantee is gone |
| Sensor | Observation — records that something was taken | The log becomes an assertion rather than a measurement |
The lock is also what the quality system depends on. IATF 16949 requires error-proofing devices to be verified on a schedule with a reaction plan for failure — meaningless for a lamp — and ISO 13849-1 governs the interlock circuit itself, because a bin lock with no safety-rated path is a lock whose failure mode is to open. The indicator and confirmation hardware are sized for oil mist, vibration and dust, since a fogged sensor or a binding lock quietly undoes every guarantee on this page.
Buying pick-to-light
The lamps-only specification, costed
Drop the interlocks to save money and you have bought a different product that demonstrates identically.
Worked example: 600,000 picks a year at a bay of close-cousin variants
Unaided, operators choose the wrong bin 1.4% of the time. An indicator is noticed and acted on for 92% of picks. An interlock fails at 40 parts per million and upstream stocking errors run at 30. Lamps only: 0.014 × 0.08 = 672 wrong picks a year. With interlocks: 0.00004 + 0.00003 = 42 a year — a factor of sixteen.
The lamps-only bay delivers a 92% reduction, which sounds close to the headline and is a different product: 630 of its 672 errors come from a moment of inattention rather than any component. The 42 from hardware can be verified on a schedule and trended; the 672 from attention cannot, which is why a quality system cannot rely on them.
If a proposal has no per-bin line item for the interlock and its verification, it is a proposal for a warning system. That may be a reasonable purchase. It is not the purchase that produces zero wrong picks.
Sensor confirmation
Why pick confirmation comes from a bin sensor, not a button
A button log records what the operator asserted. A sensor log records what happened.
A button confirmation is defeated by an operator advancing without picking, or pressing ahead to prime the next step — well-documented adaptations to rate pressure. A sensor confirmation is defeated only by a failed read.
Worked example: 40,000 picks a month
Under rate pressure, operators advance without taking on 0.3% of picks and confirm ahead of the action on another 0.8%; a sensor misses 0.05%. The button log contains about 440 confirmed picks a month that never happened. The sensor log misses about 20 that did. Both look complete, with operator, bin, kit and timestamp on every row.
The errors run in opposite directions. A button log over-reports picks and so under-reports problems; a sensor log under-reports picks and so over-reports them. For an audit trail, erring toward reporting a problem that did not happen is the safer failure. Operators are authenticated by badge or biometric, because a record attributing every pick to whoever last logged in will not survive the first dispute.
Operator ramp-up
Why operator ramp-up collapses rather than shortens
The hard part of kitting close-cousin variants is discrimination, and discrimination can be deleted.
Training load grows with the number of variant pairs an operator must tell apart, and with how similar each pair looks. Adding a seventh variant to six adds six new pairs, not one part. Kits resolve from ERP and MES work orders — a kit barcode starts the sequence and the station loads that unit's pick list — so no person chooses the variant at any point, there are no pairs to distinguish, and what is left to learn is reach, take, place.
If there is a dropdown anywhere, the discrimination has only moved from the bin to the screen — a close-cousin variant selected wrongly there produces a perfectly executed wrong kit, and the ramp-up result should not be expected. It also will not reproduce on a bay of visually distinct parts, whose ramp-up was never weeks to begin with.
Residual error
Zero wrong picks is not zero wrong kits: where the error budget goes next
An interlock guarantees which bin opened. Two things it cannot guarantee become the whole error budget.
residual kit errors = interlock failures + wrong quantity from the right bin + bins stocked with the wrong part
A bin stocked with a close-cousin part produces a lock opening on the correct bin, a sensor confirming a take, and a pick logged against the right part number — a complete, consistent record of a wrong kit, from a system working exactly as designed.
Worked example: 8,400 wrong kits a year, decomposed
Before deployment, 8,190 were operator bin selection, 150 wrong quantity and 60 wrong stocking. The interlock removes the first term: roughly 190 remain — a 97.7% reduction. But the composition inverts: before, 97.5% of errors were selection; after, essentially all are everything else, and quantity error alone is 79% of what remains.
A plant reading only the headline concludes the problem is solved. A plant reading the composition installs quantity sensing next, and makes replenishment its own error-proofed station, resolved against the same work orders — neither visible as a priority until the larger term was gone.
Field evidence
The published deployment, and the three caveats that belong with it
One of the clearest single-product attributions in MileSoft's portfolio.
A Tier-1 automotive supplier building radar modules ships six close-cousin variants with small visual differences and very different downstream calibration. Its operator pick-error rate ran at about 1.4%, and training a new operator to hold cycle time on more than two variants took four to six weeks. After this module was installed on every kitting bay, the deployment reports zero wrong-part picks since rollout and operator ramp-up of five days. Both are attributed to this module, and the published mechanisms are the two above: the interlock, and variants resolved from the work order.
Zero is a count, not a rate
It covers an unstated period and volume, so no error rate can be computed from it. Carry "over 90%" into a new site, not zero.
One site, a confusable family
The 1.4% baseline belongs to an unusually confusable variant family; the ramp-up result specifically will not reproduce where variants look different.
Picks, not kits
Zero wrong-part picks is not zero wrong kits, for the residual reasons above.
The same deployment's 92% reduction in customer containment incidents belongs to the ADAS radar assembly alignment station, and its 18% throughput gain to the programme as a whole; neither is claimed here. The account is in the pokayoke case study.
Evaluating pick-to-light
Eight questions to put to any pick-to-light system
Each is answerable by demonstration — including by us.
| # | Dimension | Ask the vendor to demonstrate |
|---|---|---|
| D1 | Control not warning | Reach into an unlit bin. Does it open? |
| D2 | Interlock integrity | What is the lock's failure mode — fail closed, or fail open? |
| D3 | Sensor confirmation | Press confirm without taking anything. Does the station advance? |
| D4 | Variant from the order | Does any person choose which kit is being built, at any point? |
| D5 | Verification schedule | How is each interlock proven still working, and how often? |
| D6 | Attributable events | Is the operator on a pick record authenticated, or inherited from a login? |
| D7 | Quantity coverage | Take three when the display says two. Does anything notice? |
| D8 | Replenishment coverage | How does the system know the bin holds what it thinks it holds? |
D1 takes three seconds and settles the category. D8 is the one nobody asks, and the one that dominates the error budget after deployment.
Standards referenced: ISO 13849-1, IATF 16949, ISO 22400-2, IEC 62264-3, IEC 62541 (OPC UA), ISO/IEC 19987, GS1 General Specifications, ISO 5725-1. Derivations are in the pick-to-light method paper; for the poka-yoke argument in plain terms, see pick-to-light, explained.
Features
Key Features
Built-in capabilities that set our software apart
Light-Directed Picking
Only the correct bin illuminates for the active kit, with quantity shown at the pick face
Bin Interlocks
Adjacent bins stay locked until the right pick is confirmed, so a wrong pick is prevented rather than flagged
Sensor-Confirmed Picks
Confirmation comes from a bin sensor, not a button press, so the record reflects what was taken
Full Pick Logging
Every pick captured with operator, bin, kit, and timestamp for audit playback
Work-Order Integration
Kits resolve from ERP and MES work orders, so the operator never selects the variant
Multi-Variant Handling
Close-cousin variants are distinguished by the system rather than by operator memory
Benefits
Key Benefits
What your team gains from adopting our platform
Testimonials
What Our Clients Say
Trusted by industry leaders across manufacturing, HVAC, and logistics.
"Our “Part-Process Matrix” as a concept and Software received 1st runner-up award in MVML Innovation fair 2014. As a customer we appreciate your support and efforts to deliver it on time and most importantly as per customer requirement."
FAQ
Pick-to-Light Software — FAQs
Answers to the most common questions about this product. Need more? Book a demo or reach out via Contact.
Is the light itself the error-proofing?
How is this different from scan-to-verify?
How quickly do operators reach full productivity?
Can it drive from ERP work orders?
What happens when an operator deviates?
Is pick-to-light a poka-yoke system?
Explore More
Related Products
Complementary tools teams often pair with this solution
Related Reading
Go deeper
Explore the long-form story behind this product.
The Light Is Not the Poka-Yoke: Interlocks, Sensor Confirmation, and Where the Error-Proofing Actually Lives
Architecture, Computational Methods, and Field Evidence from MileSoft Pick-to-Light
Read the paperRead the full Pick-to-Light Software for Your WMS or ERP deep-dive
Read articleSee how MileSoft delivers in Pokayoke
Read case studySee your savings in under a minute
Estimate the time and cost this tool could give back — free, instant, no sign-up.
Reach into an unlit bin and see what happens
Bring your most confusable variant family. We will show the interlock refusing a wrong pick, the sensor refusing a confirm without a take, and what your error budget looks like afterwards.