Ask three vendors to quote the same MES to ERP link and you will get three numbers with nothing in common. A tenfold spread between the lowest and the highest bid is normal in this market. Yet MES integration cost has little to do with which MES or which ERP you run. It is driven by the number of interfaces you have to build and by what those interfaces talk to.
We spend our days engineering industrial IoT and manufacturing platforms on Azure. Our day rates are public: 150, 200, 300, 400 and 450 euros per day depending on the profile. This article lays out the method we use to estimate MES integration cost in engineering days, with the full arithmetic on that grid. It is written for the operations or IT director of a European plant who has to defend the number internally.
Why MES integration cost varies by 10x
The software is not the main cost line. An interface is. Every machine you connect, every protocol you speak, every ERP object you synchronize is an interface that has to be specified, built, tested and maintained.
A line with three OPC UA machines and two ERP flows is five interfaces. A plant with four lines, three generations of PLCs, quality flows and lot traceability is forty. Same MES in both cases. Completely different budget.
That is where the quote spread comes from: the vendor prices its software, the integrator prices your interfaces. Until someone has counted them, every quote is a bet.
Step one: the interface inventory
Before any serious number, you count. Four columns are enough:
- Machines and PLCs: brand, year, available protocol (OPC UA, Modbus TCP, a proprietary serial link, files dropped on a network share).
- ERP objects in scope: production orders, items, bills of material, production declarations, stock movements.
- Quality flows: in-line checks, non-conformities, lot blocking.
- For each flow: direction, frequency, volume, and what happens when it stops.
A proper inventory on one line takes five to ten days of a senior engineer. At 400 euros per day, that is 2,000 to 4,000 euros. It is the best money in the whole project: an interface discovered during the inventory costs a day of analysis, the same interface discovered during commissioning costs a downtime window.
The architecture choice sets the budget
Three families, three cost profiles.
A point-to-point setup wires each machine straight into the ERP. Unbeatable up to three or four interfaces, unmanageable beyond that: every addition multiplies the connections you have to maintain.
A message broker (an MQTT server on the shop floor, Azure Service Bus on the ERP side) decouples the two worlds. The foundation costs a few extra days upfront and pays for itself around the fifth interface.
A full ingestion layer takes over once you have several lines, history and analytics: Azure IoT Hub for collection, Azure Functions for transformation, Service Bus feeding the ERP, storage for the historian. That is the architecture we work on every day. It is too heavy for a single line and the right call at site scale.
Our default recommendation for a mid-size plant: a broker from the first project, an ingestion layer when the second line arrives.
Where the engineering days actually go
Here is a typical first phase on a single line, priced on our public grid:
- Interface inventory and specification: 8 days at 400 euros = 3,200 euros
- Architecture and broker foundation: 5 days at 450 euros = 2,250 euros
- Machine connectors, OPC UA plus one proprietary protocol: 25 days at 300 euros = 7,500 euros
- ERP mapping, orders down and declarations up: 12 days at 300 euros = 3,600 euros
- Testing, downtime window and go-live: 8 days at 400 euros = 3,200 euros
Total: 58 days and 19,750 euros. Two thirds of the budget goes into connectors and mapping, not architecture. That is healthy: it is where your plant's specifics live.
For a multi-line program running over several months we switch to a pod model: a stable team at 15,000 to 40,000 euros per month depending on composition. For a first line, the day-based engagement above is enough.
The risks that blow industrial budgets
Three causes show up in almost every overrun.
Downtime windows. You do not test a connector on a line that is producing. If your plant stops twice a year, one missed window pushes the project back by months.
Legacy protocol surprises. The 1990s machine speaking an undocumented serial protocol exists in nearly every plant. Budget a reserve for reverse engineering, or keep that machine out of phase one.
Master data quality. Item codes that differ between MES and ERP, outdated bills of material, inconsistent units. The interface works, the data is wrong. Master data cleanup is a project of its own, so price it separately.
Red flags in a vendor's integration proposal
Integration is included in the license.
That sentence deserves one question: included for which interfaces? Other signals that should make you reread the quote:
- No interface inventory in the annex.
- A "standard" ERP connector with no list of the objects it actually covers.
- Integration days calculated as a percentage of the license price.
- No mention of downtime windows or of a rollback plan if the cutover fails.
- Master data migration declared as included, with no effort figure behind it.
A vendor sells its product, that is its job. The interface estimate has to come from whoever will build the interfaces.
A realistic first phase before committing the plant
Do not sign a site-wide rollout on a quote. A sane first phase fits on one line: two or three machines, two ERP flows (orders going down, declarations coming up), one simple dashboard. That is the 58-day scope above.
At the end of the phase you hold a connected line producing trustworthy data, a foundation you reuse on the next lines, and real numbers to decide what comes next. Your CFO will always prefer extending something that works over funding a promise.
On our side, we run a hybrid model: steering, quality and contracts in Belgium and Switzerland, close to the plants, engineering in Rabat. More than 17 projects shipped for more than 10 clients. The earliest we can start is two weeks after your request.
What this method does not cover
Some honest limits. MES and ERP licenses are not in these numbers, and neither is hardware: gateways, sensors, shop floor networking. Your own internal days are not either: production, quality and IT will spend real time on this project even though it never shows up on an invoice. And an estimate stays an estimate until the interface inventory is done. That is what turns a range into a commitment.
If you want to run your own scope through this method, price it on jadev-corp.com/quote. Describe your lines and your flows, get an instant estimate on the same rate grid, and an engineer reviews every formal quotation within one business day.