One number comes off a shift report, a whiteboard and a spreadsheet that gets reconciled Monday morning. The other comes from the operator who was standing there. Neither one comes from the machine. That is a fixable problem, and fixing it does not require replacing anything you run today.
This is not a reporting problem and it is not a people problem. Both numbers are assembled by honest people doing their best with what they can see. The trouble is that what they can see is a reconstruction, not a measurement.
The number in the board deck
The number the floor believes
Downtime gets logged at the end of the shift from memory. A 40-minute changeover becomes a round 30. A string of six-minute jams never gets written down at all, because nobody is going to stop and fill out a form six times an hour.
By the time the number reaches a dashboard, the shift that produced it has gone home. You are managing yesterday. The decision that mattered had a 20-minute window and it closed before the data moved.
The machines already know. Cycle counts, fault codes, state changes, drive loads, alarm stacks. Most of it lands in a historian or a PLC data table and stops there, behind the OT network, with nothing built to carry it across.
Planned downtime in one plant is unplanned in another. Ideal cycle time was set in 2011. Rate loss is quietly folded into availability. Four plants report OEE four different ways and corporate averages them anyway.
The cost nobody puts on the slide
Every improvement argument turns into a debate about whose number is right instead of a decision about what to fix.
That is the real loss. Not the reporting error — the quarters spent relitigating a baseline. The capital request that stalls because finance does not trust the before number. The kaizen that gets credited to a shift change instead of the fix. When the measurement is credible, those arguments end in a meeting instead of a fiscal year.
Some of it is stranded. Some of it was never created in the first place. The two problems look identical from a conference room and require completely different trades to solve, which is exactly why so many visibility projects die halfway.
Never instrumented
A 1990s press with no network port. A manual station with a foot pedal and no counter. A changeover that exists only as a line on a clipboard. No software will read a signal that was never generated. Somebody has to put a sensor on it, pull wire, and make the PLC publish the tag.
Instrumented but stranded
Cycle counts, fault codes and state changes sitting in a historian or a controller, inside an OT network that was deliberately built not to talk to the enterprise. Getting it across safely is an architecture question, not an export button.
Out but meaningless
A tag named MTR_3_RUN with a timestamp tells you nothing. It becomes useful when it is joined to the work order, the part number, the crew, the material lot and the standard cost. That join lives in the enterprise systems, not on the floor.
The central claim
Where the integration firm stops
At the MES boundary.
They can model your enterprise beautifully. They cannot specify a sensor, terminate a wire, or touch PLC logic in a running plant. So the project ends at “give us a feed” — and the feed does not exist.
Where the controls shop stops
At the historian.
They will instrument anything you point at and build a first-rate screen. Then the data needs to meet the ERP work order and the standard cost, and that is somebody else's scope. You own the gap in between.
Where we stop
At the number you can defend.
Controls engineers, integration architects and operations strategists are all our employees, on one P&L, with no layer subcontracted. The division boundary that defeats everyone else is an internal hallway here.
We are the only company in the Americas with OT engineering, enterprise integration, bespoke application development, AI development and operational strategy under one P&L without subcontracting any layer. On a visibility project that is not a credential. It is the difference between a working number and a change order.
Order matters. Every failed visibility project we get called in to rescue started at step three, bought a screen, and then discovered there was nothing credible to put on it.
Magic Engineering
Controls engineers walk the line and decide what has to be instrumented: counters, state feedback, fault capture, drive loads, vision checks, the changeover signal nobody ever wired. We pull the wire, program the PLC, and make the equipment publish honest tags. 1,900+ automation projects, 30+ years of controls work, CSIA-certified since 2002.
Magic XPI
Magic XPI is enterprise integration across ERP, CRM, WMS, MES, HR, finance, support, historians and the plant floor, with 100+ pre-built and certified connectors and proprietary in-memory middleware for high availability. It carries machine events out of the OT network and joins them to the work order, the part, the crew, the material lot and the cost. A tag becomes a fact about the business.
Magic FactoryEye
Magic FactoryEye reads contextualized data via Magic XPI and surfaces OEE, performance efficiency, cycle time and downtime in one place. It overlays the systems you already run rather than replacing them, so the number on the screen traces back to a machine event an operator can recognize.
Magic Consulting
A dashboard is not an improvement. Our operations strategists run Assess, Target, Execute, Sustain: find the constraint that actually limits throughput, tie activity-based costing to real production data, and sequence the work. Typical gains are 10-15%+ in throughput and yield.
Magic Engineering, Magic Integration and Magic Consulting are divisions of Magic Software Americas, under one P&L, with no layer subcontracted. Same project plan, same accountability, one number to call when the tag on the screen looks wrong. Nobody gets to tell you the sensor crew and the integration team are different companies, because they are not.
The test of a floor metric is simple: can the operator standing at the machine look at it and agree? These come from what the equipment actually did, so the answer is yes — and the argument about whose number is right is over.
Availability, performance and quality computed from machine state, not from a form. When the number moves you can click into the shift, the machine and the event that moved it.
Actual rate against validated ideal cycle time, per part and per line. Rate loss stops hiding inside availability, which is where most of the missing OEE has been all along.
Measured per unit and per station, so you can see which station sets the beat and what a changeover really costs you compared to the standard everyone quotes.
Captured from fault codes and state changes as it happens, attributed to the machine rather than reconstructed at the end of the shift from memory.
The six-minute jams nobody logs, counted. On most lines this is the single largest unexplained gap between the board number and the floor number.
Where the instrumentation supports it. Vibration, temperature, motor current and cycle-count trending that flags a component before it takes the line down. If the sensor is not there yet, Magic Engineering can install it.
Shift-lagged
You find out Tuesday that Monday went badly.
Useful for a trend chart. Useless for the 20-minute window in which somebody could have done something about it.
Real-time
The supervisor sees the drift while the shift is still running.
Same data, early enough to act on. That is the entire difference between a reporting system and an operating system.
If you have been quoted a multi-year platform migration to find out your real OEE, you were sold a platform, not an answer. You do not need a new ERP, a new historian or a new control standard to measure your own equipment honestly. You need the layer that connects what you have.
Being straight with you
We would rather tell you that than find out you learned it later. What we are selling you is not a product badge — it is the architecture underneath it, and every layer of that architecture is mature and in production at scale:
Judge us on the wiring, the integration and the operations thinking. The screen is the easy part — and it is the only part anybody else wanted to sell you.
Visibility work earns its keep fast, because the first thing a trustworthy number does is end an argument that has been eating quarters.
4-12 weeks
Typical FactoryEye ROI, per plant
18 plants
Deployed in 18 weeks (confidential client)
1,900+
Automation projects delivered
Since 2002
CSIA-certified controls practice
Multi-vendor coverage
Your plants were not standardized and nobody is going to let you standardize them before you are allowed to measure them. We work across the platforms you already have, and we are an Inductive Automation Premier Integrator on the Ignition side.
Institutional endorsement
CESMII CEO John Dyck described the Magic-Axiom collaboration as “the kind of partnership needed to drive the industry forward,” citing its potential to bring scalable and sustainable innovation to U.S. manufacturing.
CESMII is the U.S. Department of Energy-backed Smart Manufacturing Institute. Source: Magic Software press release, July 15, 2025.
A federally backed manufacturing institute does not hand that language out for a dashboard. It is a comment on the model: instrumenting the physical layer and connecting it to the enterprise, at a scale that works plant after plant rather than once as a pilot.
Divisions of one company, under one P&L, with no subcontracting. You are not integrating three vendors on your own time.
The physical layer
Controls engineers who instrument equipment, program PLCs, build HMI and SCADA, and run a Managed OT Platform. CSIA-certified since 2002, 1,900+ automation projects, multi-vendor across the major control platforms. This is the division that creates data where none existed.
See the engineering practiceThe data layer
Integration architects running Magic XPI across ERP, CRM, WMS, MES, HR, finance, support, historians and the plant floor. 100+ pre-built certified connectors and in-memory middleware for high availability. This is the division that gets the data out and makes it mean something.
See the integration practiceThe decision layer
Operations strategists who take a trustworthy number and turn it into throughput. Assess, Target, Execute, Sustain, with activity-based costing tied to real production data and typical gains of 10-15%+ in throughput and yield.
See the consulting practiceThe platform
How the overlay works, what it reads via Magic XPI, and what deployment looks like when you keep the systems you already run.
Read moreThe sibling outcome
Trustworthy floor data is what makes a cost number real. Once machine time, scrap and downtime are measured, activity-based costing stops being an estimate and cost genealogy becomes possible.
Read moreWhat comes next
Models are only as good as the signal underneath them. Instrumented, contextualized production data is the prerequisite nobody tells you about before the pilot.
Read moreTell us one line, one cell or one plant where the reported number and the floor number do not match. We will walk through what is already instrumented, what has to be added, and what it takes to get a defensible OEE out of it. Thirty minutes, no slides.
