I recommend you keep your ERP and add a thin edge layer plus targeted capture points — machine gateways, cameras, operator terminals — to collect time-stamped, lot-traceable shop-floor records. Validate and enrich them at the edge, then push into ERP via existing import templates, scheduled batches, or controlled screen automation to preserve audit trails.
- A thin edge layer collecting machine signals and operator inputs fixes most shop-floor data gaps without replacing ERP.
- Validate and time-stamp records at capture, then send them using the ERP's import template, scheduled batches, or controlled screen entry to avoid heavy ERP changes.
- Design mapping and audit fields up front; capture timestamps, batch IDs and signatures at the edge.
- Run pilots in shadow mode with clear acceptance and a kill criterion to stop projects decaying into maintenance debt.
- Replace ERP only when multiple core processes are fundamentally unsupported and bridges cannot deliver required traceability.
Signs this is happening in your business
- Three operators spend six to eight hours weekly each retyping machine reports into an ERP form.
- Traceability requires hunting emails and spreadsheets to link finished batches to raw material lots.
- Quality logs are written on paper and later scanned, so timestamps and signatures are missing from ERP records.
- Production planners see inventory lag by a day because machine events are not captured live.
- IT rejects ERP customisation requests as too risky and production teams maintain rogue spreadsheets.
- A single person owns the master spreadsheet that links shop-floor events to ERP orders.
- Camera snapshots of cuts, welds or labels sit on a shared drive and never reconcile to ERP.
If several of these are true at the same time, the problem is usually the process rather than the people running it.
What this is costing you a year
(Team hours per week x hourly cost of employees x 52) = annual payroll loss
- People doing this work4
- Hours each per week6
- Hourly cost of employees$45
- Weeks52
The annual payroll cost of four people spending six hours a week each on manual shop-floor data work
Find the number before you spend anything
Run the formula above on your three most repetitive processes. If any of them clears a year of a salary, that is the one to bring to a call.
How to automate shop-floor data without replacing systems
Keep your ERP and add a thin edge layer plus targeted capture points — machine gateways, camera capture and operator terminals. Validate and enrich records at capture with lot, operator and timestamp, then push them into ERP via its import template, scheduled batches, or controlled screen automation to preserve auditability.
- Collect data at the machine or operator point of action.
- Validate and enrich the record with lot, operator and timestamp.
- Push into ERP using import templates, scheduled batches, or controlled screen automation.
What patterns capture shop-floor data without changing your ERP
Four practical capture patterns keep ERP unchanged: edge gateways that read machine signals, periodic polling for older controllers, camera capture for labels and visual checks, and operator terminals or mobile forms for manual inputs. Pick the pattern by the exact data you need and the interfaces your machines expose.
- Edge gateway beside machines to read signals and produce timestamped records (OPC-UA, MQTT where available).
- Periodic controller polling for older PLCs to pull counters and status on a schedule.
- Camera capture to read labels, verify assembly steps, or attach images to lot records.
- Operator terminals or mobile forms for job selection, lot entry, pass/fail and signed acknowledgement.
Where the money is lost now and what automation buys back
Payroll drains come from retyping, chasing missing signatures, and reconciling spreadsheets. Automation captures records at the point of action, enforces validation, and routes exceptions to humans. That reduces hours spent on data entry, shortens audit assembly time, and shifts staff to managing exceptions rather than copying data.
If your team treats a spreadsheet as the master record, automation will only paper over a governance hole until an exception breaks traceability.
Cost of the manual process across a team
Compute the true annual payroll cost by multiplying the number of people, hours per week, hourly pay and weeks per year. Use that single figure to compare against a fixed-scope integration or supplier quote so stakeholders can see the payback in plain terms.
Data mapping, validation and audit fields you must include
Capture who performed the action, which lot it belongs to, and when it happened at the moment of action. Make required fields and formats explicit, record signed acknowledgements, and implement write-once change logs in the capture layer so incomplete or altered records never reach ERP untracked.
- Mandatory fields: lot/batch ID, machine ID, operator ID, start/finish timestamps, quantity produced, non-conformance code if any.
- Signatures: capture a signed acknowledgement or operator PIN and store the image or hashed token with the record.
- Audit trail: keep a write-once record of changes, who made them, and why, stored with the captured data.
How to push shop-floor records into ERP without heavy ERP changes
Use the ERP capability already present: scheduled file imports if it supports templates, live event feeds where available, or controlled screen automation where no import exists. Always choose the least invasive path that preserves upgradeability and avoids creating bespoke fields inside ERP.
- Scheduled imports: create hourly or daily files that match the ERP import template.
- Live connections: send single confirmed records when ERP accepts live events.
- Screen-based automation: a controlled routine enters fields into the ERP UI with logging and retries.
Automating a broken process gives you speed but not reliability; fix rules and ownership before automation.
Examples mapped to common ERP behaviours
Match your shop-floor source to the ERP's accepted input. Schedule imports for ERPs with batch templates, send single events where live updates exist, and fall back to controlled screen entry where imports are absent. Prefer formats the ERP already accepts to avoid upgrade complications.
- If the ERP has a production-report import, schedule hourly imports from the edge collector.
- If the ERP accepts live inventory adjustments, send confirmed completion events as they occur.
- If the ERP has no import facility, use controlled screen automation and capture success logs.
Build, buy or hire: practical trade-offs
Buy a packaged product when you need speed and the product models your process closely. Build when your process is unique and you can own maintenance. Hire for a fixed-scope delivery when you lack internal technical leadership but require a documented handover and predictable delivery.
- Buy: fast time to value, may require changing processes to fit the product.
- Build: tailored to your process, needs someone to own it long term.
- Hire: fixed-scope delivery with training and documented handover.
Identify your biggest operational bottleneck
Most owners already know which process is the problem. What they do not have is the annual figure attached to it. That is a 15-minute conversation, not a project.
A rollout playbook that avoids the pilot that dies
Run pilots in shadow mode with clear acceptance criteria and a kill criterion. Shadow mode means automation produces records while humans continue entering data; compare outputs for completeness, latency and error rate. Handover only after meeting criteria and providing runbooks, training and a support window.
- Define acceptance: data completeness, time-to-record, and acceptable error rates.
- Run in shadow mode for a defined period and compare machine and human records.
- Handover with runbooks, training sessions, and a 30–90 day support window.
Replacing ERP to fix missing shop-floor data is rarely the correct first step; bridges are usually cheaper and faster.
Common failure modes and the sequence that kills projects
Failure is usually organisational: a pilot shows partial success, nobody defines done, no operations budget is set, and the automation runs without an owner until spreadsheets reclaim control. Stop that by naming a single accountable owner, codifying acceptance, and funding operations before switching from shadow to live.
- Failure sequence: partial pilot → no acceptance criteria → no operations budget → automation drifts → manual work returns.
- Operational signs: daily supplier fixes, no internal owner, and spreadsheets regaining authority.
When replacing ERP is actually the right decision
Replace ERP only when core business processes are fundamentally unsupported and every bridge forces repeated custom work that costs more than replacement. If three or more major processes need constant ERP customisation or you cannot meet regulatory traceability through any bridge, replacement becomes defensible.
- Replace ERP if multiple modules fail regulatory needs and bridges always require custom work.
- Compare replacement cost against a two-year plan of bridges and targeted upgrades.
Examples and timelines for common integration types
Costs vary with wiring and machine interfaces. Typical packages include a small terminal plus CSV import, an edge gateway for a single line, and a full-plant rollout with cameras and screen automation. Measure engineering days and licence needs to compare bids rather than rely on headline prices.
- Small: operator terminals and CSV import template. Timeline: 4–8 weeks. Delivery: validated records and scheduled imports.
- Medium: edge gateway on one production line with machine signals and operator terminals. Timeline: 8–16 weeks. Delivery: live records and scheduled imports.
- Hard: full plant rollout with cameras, legacy PLCs, and screen automation for ERPs without imports. Timeline: 4–6 months. Delivery: complete traceability and handover.
Handover, support and who owns maintenance
Decide operational ownership before go-live. I recommend production operations own first-line maintenance with a named technical escalation to IT or the supplier for the initial support window. Require documented runbooks, change logs and a warranty period where the supplier fixes issues found under real load.
- Owner: production operations handles first-line support.
- Escalation: a named technical contact handles configuration and network issues.
- Documentation: runbooks, change log and a shared knowledge base for the frontline team.
What to do tomorrow to make progress
Run one focused technical review that maps a single process end-to-end; it gives a recommended architecture and a fixed-scope estimate you can budget. I run those reviews and produce a short handover plan — see How to work together to book a 30-minute technical review.
Where businesses waste the most time
- Re-keying machine outputOperators spend hours each week retyping counters, causing payroll waste and delayed recordsInstall a small gateway that reads machine signals and writes hourly import files matching the ERP template, keeping the ERP unchangedA person still checks: Operators confirm job and quantity and approve the hourly file when exceptions occur
- Chasing paper signaturesQuality staff assemble and scan paper logs for audits, often with missing timestampsUse terminals for operator sign-off at the point of action and attach the signed image to the record before importing to ERPA person still checks: Quality manager reviews flagged non-conformances and approves overrides
- Spreadsheet master filesA single spreadsheet becomes the truth, blocking visibility and creating a single point of failureCapture the same fields at the operator terminal and feed them into ERP imports so ERP remains the master recordA person still checks: Process owner reconciles spreadsheet exceptions and updates mappings when processes change
How this compares with what you already run
| Operational area | The manual way | The automated way | Annual business impact |
|---|---|---|---|
| Production reporting | Operators email or retype counts at shift end | Automatic capture and hourly import to ERP | Saves labour equal to one full-time salary |
| Traceability | Paper logs and spreadsheets searched after an event | Records include lot ID and timestamp at capture | Reduces audit assembly time and customer rework |
| Quality sign-off | Paper signatures filed separately | Signature captured at terminal and stored with record | Prevents regulatory rework and late corrections |
| Inventory updates | Planners wait for end-of-day reports | Confirmed completions update ERP hourly | Fewer stockouts and lower expedited shipping costs |
| Label verification | Photos on a shared drive, not reconciled | Camera captures label and links to lot record | Reduces mislabel incidents and credit claims |
| Exception handling | Ad-hoc emails and calls to resolve rejects | Exceptions logged automatically and routed | Faster resolution and lower scrap costs |
What the end-to-end workflow looks like
- TriggerMachine finishes a run or an operator completes a step and submits a recordEdge collector captures the event and stamps it with time and lot
- EnrichmentThe record is enriched with operator ID, shift and material lot numbersA rules routine validates required fields and flags missing data
- Image captureIf needed, a camera takes a photo of the part or label and links it to the recordImages are stored and referenced by the record
- ValidationThe record is checked against business rules for acceptable values and rangesAutomated checks mark the record good or move it to an exception queue
- ExportValidated records are prepared in the ERP import format or queued for UI entryScheduled import files are created and uploaded, or the screen routine starts to enter records
- ERP confirmationERP accepts the import and returns a confirmation or error listImport status is logged and errors are routed to a person
- Exception handlingA human reviews flagged records, corrects data and resubmitsStays human
- Audit storageFinal records, images and signature tokens are archived for regulatory reviewThe system stores a write-once record and the change log
The metrics that tell you it worked
| Metric | What it tells you |
|---|---|
| Hours reclaimed per week | Total operator hours saved by removing manual rekeying; measures direct payroll savings |
| Time to ERP record | Time between an event on the floor and its arrival in ERP; measures visibility improvements |
| Number of missing signatures | Count of production records missing a required signature; measures compliance risk reduction |
| Exception rate | Proportion of records flagged for manual review after automated validation; measures data quality |
| Audit retrieval time | Hours required to assemble a batch trace for an audit; measures traceability improvement |
| Rework incidents per month | Count of batches returned for rework due to data or label errors; measures impact on quality |
When this is the wrong move
- You are planning a major business model change that requires a different ERP anyway.
- All shop-floor processes require complex ERP logic that cannot be represented in import files or entry routines.
- You lack any internal owner and will not allocate headcount to operate or approve exceptions after go-live.
- Your ERP is under contractually imminent replacement within six months and the bridge would be discarded immediately.
- Your machines are legacy and cannot be read or are behind proprietary controls with no feasible adapter.
The mistakes that cost the most
- Pilot without acceptance criteriaTeams run a pilot, see partial success, then nobody defines pass/fail; the pilot continues indefinitely and never becomes production.
- Automating a broken processAutomation replicates bad paper rules: automation runs → bad records flood ERP → quality flags increase and corrections become costlier.
- No operational ownershipSupplier builds and supports temporarily, then responsibility lapses: handover with no named owner → small issues accumulate → staff revert to spreadsheets.
Frequently asked questions
Will this approach meet regulatory audit requirements?
Yes if you capture required fields at the point of action and store a tamper-evident audit trail with timestamps and signatures. Implement those controls in the capture layer and archive an immutable record alongside the ERP import so auditors can recreate the timeline without relying on paper.
How long does a typical pilot take before we see results?
A focused pilot on one line or process can deliver measurable results in 4–8 weeks. That includes defining required fields, installing capture devices or terminals, and running shadow mode to compare automated records with the current manual process.
What if our ERP vendor refuses imports?
If the ERP truly offers no import path, controlled screen automation that enters records like a person is a pragmatic fallback. It is slower than direct imports but preserves the ERP and avoids risky custom changes.
Do we need to buy new hardware for every machine?
Not always. Some older machines can be polled with a small adapter; others are best handled with operator terminals or periodic polling. A single gateway can often serve several machines on one line.
Should I buy a packaged product or hire someone to build this?
Buy if you prioritise speed and the product models your process closely. Hire for a fixed-scope build when processes are unique and you want documented handover. I recommend a short advisory engagement to map one process first; see Services and How to work together for options.
Stop paying salary for robotic work
Bring your messiest process to a 15-minute audit and we will map where the hours actually go, live on screen. You leave with the number whether or not you work with me.