01
Edge gateways read the plant once
PLCs, SCADA platforms, historians and plant systems are connected through secure edge gateways and validated data interfaces such as OPC UA and MQTT. The control logic that runs production is not modified.
MES software development and migration
The MES still runs the plant, but nobody wants to touch it.
Migrate off a legacy MES one work center at a time. The old system keeps running until the new one proves itself.
A manufacturing execution system cannot be switched over on a weekend. We build the new MES beside the one you run, feed both the same PLC, SCADA and historian data, compare what they produce, and move production onto the new system work center by work center, with a way back at every step.
- Parallel run before any switch-over. New and legacy systems run on real production data at the same time, and equivalence checks compare their outputs before anything depends on the new one.
- No rip-and-replace. Legacy PLCs, SCADA platforms and historians are connected through secure edge gateways and validated interfaces such as OPC UA and MQTT, without touching the control logic that keeps the line running.
- Cutover is controlled per work center and per shift, so a problem affects one station on one shift, not the plant.
- Capped packages of one to three weeks. You pay actual effort up to the cap, never above it, and you can pause at any package boundary with production running.
5.0Clutch rating, 45 reviews
100+certified engineers
12+industries served
98%client recommendation rate
Results from Devox manufacturing projects
Measured on real plants, not in a demo
70% OEEachieved with a custom MES and Siemens IoT sensorsNorthern European automotive manufacturer
−15%cost losses from scrap and low qualityMulti-brand automotive production
−25%supply chain disruptionMulti-line production across contractors and time zones
30%faster issue resolution cycleMulti-site quality management system, five plants
How the migration runs
The new MES is built beside the old one, not on top of it
Both systems read the same machine data. The new one earns each work center by producing the same numbers as the system you trust, on real production, before anyone depends on it.
02
Shadow run on real production data
The new MES processes the same orders, counts and events as the legacy one and writes its results to its own store. Nothing on the floor changes while this runs.
03
Equivalence checks decide the cutover
Outputs of old and new are compared on real data: order status, counts, scrap, downtime reasons, OEE. A work center moves only once its numbers match and stay matched.
04
A way back on every step
Which work center runs on which system is configuration, not a deployment. A station can be returned to the legacy MES between shifts while the cause is found.
Cutover state, in practice
Line 1 / Assembly : new // migrated, 6 weeks in production Line 1 / Paint : shadow // both systems running, outputs compared Line 2 / * : legacy // untouched, not scheduled yet
Every station sits in one of three states, and moving between them is a configuration change with a defined rollback, not a plant-wide release.
Two ways to replace an MES
A weekend cutover puts the whole plant on one untested system
Both routes end with the legacy MES retired. They differ in what a mistake costs while you get there.
Rip and replace
Everything moves on one date
- ✕The first real test of the new MES is the plant depending on it.
- ✕Rollback means restoring the old system across every line at once.
- ✕Operators learn the new terminals during a live ramp-up.
- ✕Data gaps show up as missing traceability records after the fact.
- ✕The estimate is fixed before anyone has seen how the plant's data actually behaves.
Staged MES migration, Devox
One work center at a time
- ✓The new MES runs in shadow on real production data before it owns anything.
- ✓A station returns to the legacy MES through configuration, between shifts.
- ✓Operators are trained on one station while the rest of the plant works as usual.
- ✓Traceability is reconciled against the legacy records before the cutover, not after.
- ✓Each package is priced from the pace measured on your actual systems and data.
Plan the first work center 30 minutes on Zoom, no cost
The first package
A two-week foundation stage that ends with real data flowing
Everything after it runs on this machinery, and it produces the plan and the estimates for the next packages from your actual systems rather than from assumptions.
Week 1
Audit and estimation plan
A solution architect maps the current MES, the interfaces it depends on, the data model and the workflows that must survive the move, with your engineer in a walkthrough session.
You provide: system access, and same-day answers on how the plant really works.
Week 1–2
Data path and environments
Edge gateway and interface setup for one work center, the delivery pipeline, environments, access control and the rollback path, plus the reconciliation checks that will decide any cutover.
Roles: senior engineer and DevOps, with a PM as your single point of contact.
End of week 2
One work center in shadow
The new system consumes live data from that station and produces its own counts and statuses beside the legacy MES, with the first comparison report on real production.
You get: a working data path, a comparison you can read, and the estimates for the next packages.
How the work is contracted
Capped packages instead of one padded number
Work is contracted in short packages of one to three weeks. A larger milestone, such as a full line, spans several of them.
Capped work packages
- ✓Each cap is your worst case. You pay actual effort up to it, never above.
- ✓When the work outruns the estimate, the package closes under cap and you keep the difference.
- ✓The next package is priced from the real pace measured on your systems, so proven speed makes every following quote cheaper.
- ✓Each package ships complete and reversible, so you can pause at any boundary with production running.
One fixed price for the whole replacement
- ✕Padding for uncertainty is baked into the number on day one.
- ✕When delivery gets faster, that padding stays with the vendor and the gains never reach your invoice.
- ✕No risk above the cap and all upside below it is exactly what this shape cannot offer.
- ✕Pausing mid-way leaves the plant split between two systems with no working boundary to stop at.
Two things are decided after the foundation stage rather than guessed now: the cutover order across work centers and lines, and how much of the integration work your automation team keeps in house with our architecture support.
Scope
What the new MES has to keep doing on day one
A migration is judged on what does not break. These are the functions we reconcile against the legacy system before a work center moves.
Execution
- Production orders and dispatching
- Work instructions at the station
- Operator terminals and sign-offs
- Material consumption and genealogy
- Schedule adherence
Measurement
- OEE and its drivers: availability, performance, quality
- Downtime capture and reason codes
- Scrap and rework
- Throughput and cycle times
- Shift and line reporting
Integration
- ERP: orders, confirmations, inventory movements
- SCADA, PLCs and historians through edge gateways
- QMS and quality checks
- WMS and material supply
- CMMS and maintenance triggers
AI acceleration, per task
Measured per task, with a named human checkpoint on each one
Blanket discounts on AI do not survive contact with a plant. Overall project gains land around 10 to 30 percent even with near-universal tool adoption, while specific task types genuinely run twice as fast.
| Work type | AI impact | What the human keeps |
|---|---|---|
| Data models, interface definitions, scaffolding | 60–80% faster | Correctness review against the tag list |
| Converting framework-free calculation logic | 70–90% faster | Parity sign-off against the legacy results |
| Standard screens and reports | 50–60% faster | Visual QA with the people who use them |
| Operator terminal components | 40–50% faster | Usability on the floor, gloves and all |
| Scheduling, genealogy and other complex flows | ~30% faster | Business-logic validation |
| Verifying the line behaves identically | ~0% | All of it |
Why the total does not drop by 70 percent: those percentages apply to writing code. Every migrated function still needs an engineer to review the output, compare it against the running system, test the real flow, deploy and watch it. Where an automatic verdict exists, such as a reconciliation check against the legacy MES, that comparison can run continuously and the work moves at machine pace. Where the verdict is a person watching a line, verification sets the clock, and no tool removes that step safely.
Technology
The full stack behind an MES migration
From the tag on the machine to the report the plant manager opens on Monday.
Plant floor and connectivity
OPC UAMQTTModbusSiemens PLC and IoT sensorsSCADA platformsProcess historiansSecure edge gatewaysStore and forward buffering
MES services
.NET 8 / C#JavaNode.jsPythonREST and GraphQL APIsEvent streamingRules and workflow engineRole-based access controlAudit trails
Operator and management interfaces
ReactNext.jsTypeScriptTouch and kiosk terminalsOffline-tolerant screensReal-time dashboardsShift and OEE reporting
Data
PostgreSQLMicrosoft SQL ServerTime-series storageData lake and warehouseReconciliation and diff checksRetention and traceability records
Integration with the systems you keep
ERP: SAP and customWMSQMSAPSPLMCMMSWorkflow layer beside the ERP
Delivery and operations
On-premise, private cloud or AWS and AzureDockerKubernetesCI/CD pipelinesInfrastructure as codeObservability and alertingDisaster recovery drills
Where a plant cannot send data outside its own network, the system runs entirely on premise or in a private environment, and redaction layers keep sensitive information inside approved boundaries.
Case studies
Plants that changed systems without losing a shift
Custom MES across multi-brand production
Production visibility, traceability and workflow automation across multiple lines, contractors and time zones.
- 70%overall equipment effectiveness
- −25%supply chain disruption
- −15%cost losses from scrap and low quality
Multi-site quality management system
A custom QMS for a Central European industrial-automation manufacturer, deployed to multiple sites at once.
- 30%faster issue resolution cycle
- 0downtime when new audit modules ship through CI/CD
Workflow layer beside the ERP
Tax, procurement and quality approvals moved out of spreadsheets and email without opening years of custom ERP code.
- 3d → 4haverage approval time
- 72% → 98%SLA hits for finance and quality
- 12,000workflow calls per day on existing hardware
FAQ
What plant and IT leaders ask first
Will the migration stop production?
No. New and legacy systems run in parallel on real data without affecting users, and the switch-over is gradual, work center by work center, without stopping operations or risking data loss.
Do we have to replace our PLCs, SCADA or ERP?
No. Legacy PLCs, SCADA platforms, historians and plant systems are connected through secure edge gateways and validated interfaces. The control logic is not modified, and workflow layers can sit beside the ERP rather than inside it.
How do you prove the new MES counts the same as the old one?
Equivalence checks compare the outputs of both systems on real production data: order status, counts, scrap, downtime reasons and OEE. A work center moves only after its numbers match and keep matching under normal load.
What happens to historical production and traceability data?
Historical records are migrated with backups, validation checks and controlled transfers, combining automated tooling with manual review, and reconciled against the legacy system before the old one is retired.
Can we stop after the first work center?
Yes. Each package ships complete and reversible, so you can stop at any boundary with production running on a mix of both systems and resume later.
We run several plants. How does this scale?
Once a solution is validated in a pilot, we replicate it across sites with structured release management, standardized processes and metrics, so the second plant is a rollout rather than a project.
Does our production data leave our environment?
Only if you want it to. The system can run fully on premise or in a private environment, and redaction layers prevent sensitive information from leaving approved boundaries.
Find out what moving off your MES would take
30 minutes on your plant, your current system and its interfaces: which work center moves first, what it has to prove, and what the first two-week package would deliver.
Pick a time Free call on Zoom