Why More Manufacturing Data Hasn’t Made Decisions Any Easier

A plant manager pulls up three dashboards before a 7 a.m. production meeting. One shows yesterday’s yield. One shows open work orders. One shows a maintenance alert from a machine that’s been acting up for a week. None of them tells her what to do about the material shortage that just landed in her inbox.

So, she starts calling people. Same as always.

It’s a strange place to end up, honestly. Manufacturers have never had more information than they do right now. Sensors on nearly every machine. Cloud platforms holding years of history. Dashboards for quality, throughput, downtime, scrap, you name it. Visibility into the floor has gone up a lot over the last twenty years, by any measure you’d want to use.

But the hard calls haven’t gotten any easier. Production leaders still juggle shifting customer priorities against capacity that refuses to move. Engineering still can’t launch a new product without stepping on whatever’s already running. A late shipment, a failed inspection, a machine going down, any one of those can blow up a whole day’s plan by 9 a.m., and somebody still ends up walking the floor or on the phone trying to sort it out.

More data hasn’t fixed this because data was never the problem

Most plants already have thousands of reports, alerts, and KPIs sitting around unused. What’s missing is the connective tissue between them. The relationships that explain what one piece of information actually means once you account for everything else happening at the same time.

Deloitte’s 2025 Smart Manufacturing and Operations Survey found that 92% of manufacturing executives expect smart manufacturing to drive their competitiveness over the next three years. Same research, though, points out a familiar snag. Data still sits siloed across plants, and across IT and OT systems, and that gap is one of the biggest things standing between what manufacturers have already invested and what they were hoping to get out of it.

Here’s a small example. An engineering change looks pretty much routine on paper. Update the drawing, release the revision, move on. In practice it touches work instructions, quality plans, tooling requirements, and purchasing before it’s anywhere near the shop floor. By the time it gets there, sequencing and delivery commitments have usually already shifted, whether anyone caught it or not.

Experienced people have their own way of working around this. When something goes sideways, they don’t hunt for one number in one system. They pull from wherever they can, a report here, a hallway conversation there, a gut check built on years of watching how a particular line behaves, because they already know the real answer was never sitting in one place to begin with. It’s in how the product, the process, the materials, the people, and the customer commitment all connect to each other.

Why the Systems Still Act Like Manufacturing is a Bunch of Separate Departments

There’s a fairly plain, mostly historical reason for this. Manufacturing software grew up solving one problem at a time. PLM took engineering data. ERP handled purchasing, inventory, and the financial side. MES coordinated the floor. Quality systems tracked inspections on their own schedule. Planning tools optimized capacity.

None of that was a wrong call at the time. Each system solved something genuinely useful, and honestly, each one still does. The complaint here isn’t that PLM is bad at managing engineering data or that MES is bad at coordinating the floor. It’s that manufacturing itself was never divided up into those boxes in the first place. A product moving through a plant touches engineering decisions, material movements, production, quality checks, maintenance, scheduling, customer commitments, often all within the same shift, and none of it happens on its own even though the software running underneath mostly still treats it that way.

That mismatch is a big reason digital transformation projects so often end up as a handful of small wins instead of a real gain across the whole operation. You can tighten up engineering, or production, or quality, or planning by itself and still have no idea how a decision made in one of those areas hits everything downstream of it. A plant might genuinely improve first pass yield on one line and still miss its delivery targets that same quarter, because the improvement never got connected to what it meant for scheduling or purchasing.

McKinsey found a similar pattern with AI specifically. Nearly 90% of manufacturers are experimenting with it in some form. Only about 7% have gotten it to scale past a pilot. Usually, technology isn’t what’s holding it back. The data and systems underneath were never built for anything beyond one narrow job, so the moment somebody tries to point AI at a broader question, the same fragmentation that’s always slowed people down starts slowing the algorithm down too.

The Missing Piece isn’t More Automation

More automation, more connectivity, more AI, all useful, none of them the actual point. They are tools. The real gap is that most manufacturing operations don’t have a shared, structured understanding of how their products, processes, materials, equipment, and people relate to each other, something that every system and every person could pull from instead of rebuilding it themselves, again, every single time.

Think about what happens when a machine reports “operation complete.” Technically that’s accurate, and the MES updates without a hitch. Operationally, though, that one message doesn’t say which product got built, which customer order it belongs to, which engineering revision was followed, whether the required inspections happened, or what it means for everything scheduled behind it. Somebody still has to go find that out, usually by asking around the floor.

Worth pulling apart the words here too, because manufacturing tends to blur them together. Data is just what happened. Information is that data organized enough to communicate. Understanding only shows up once information gets read against everything else going on around it at the same time, and that last part is exactly what most manufacturing systems still leave for a person to figure out. A dashboard can tell you a number moved. It can’t tell you why it moved, or whether that movement even matters given what else is going on that week.

This distinction matters more than it sounds like it should, because a lot of manufacturers genuinely believe they’ve solved the understanding problem once they’ve solved the connectivity problem. They connect MES to ERP, wire up a few APIs, get real-time data flowing between systems that used to require someone keying things in by hand, and the assumption is that understanding follows automatically once the plumbing works. It doesn’t. Two systems can exchange information perfectly and still interpret that same information in two different ways, because nobody ever defined the relationships that make the information meaningful in the first place.

What this Actually Costs a Plant

Treat it as an abstract systems problem, and it’s easy to wave off. In practice it’s real hours and real risk. Engineering, production, quality, and planning will look at the same event and end up asking completely different questions. Was the right revision built? What’s the delivery impact? Is there a quality concern buried in there somewhere? Sorting all that out eats time that should’ve gone toward actually deciding something.

It shows up in who a plant leans on, too. The people who know how to bridge these gaps, who trust which report over which, who to call when two systems disagree, are usually your most experienced folks. That knowledge is hard to write down and even harder to hand off, and every retirement takes a slice of it out the door. Plenty of plants have felt this directly. Someone retires after twenty-five years and suddenly nobody quite knows why the second shift always runs slower on a particular line, or which supplier’s material tends to cause trouble in a specific process step. That knowledge was never in a system. It was in a person’s head, and now it’s gone.

New technology inherits the same problem when it walks in. AI, advanced planning, digital twins, all of it depends on interpreting operational information correctly. Fragmented relationships underneath don’t get fixed by adding these tools on top. They just get processed faster, fragments and all. A digital twin built from an incomplete picture of the factory will confidently simulate outcomes that have nothing to do with how the real factory behaves, and it’ll do it quickly, which somehow makes the wrong answer feel more convincing rather than less.

There’s also an obscured cost that doesn’t show up on any spreadsheet. Every time a team spends an afternoon reconciling three versions of the same production event, that’s an afternoon not spent improving anything. Multiply that across every plant, every week, every year, and it adds up to an enormous amount of effort spent just getting everyone onto the same page before the actual work of getting better can even start. None of that shows up as a line item anywhere. It just shows up as a plant that always seems a little busier than it should be, relative to how much is actually getting produced.

The Same Event, Five Different Questions

It’s worth walking through how this plays out across a single day, because the abstract version of the problem undersells how often it happens. Take that same “operation complete” message from earlier. Engineering wants to know if the correct revision was followed, since a change went out two weeks ago and nobody’s fully confident it made it to every work center yet. Quality wants to know whether the required inspections happened at the right point in the process, not just whether they happened at some point. Planning wants to know what this means for capacity for the next operation, since a downstream work center is already tight this week. Maintenance is wondering whether the equipment that ran this job shows the same wear pattern as the one that failed last month. Purchasing, if anyone thought to ask, would want to know whether the materials consumed match what was actually ordered, since a substitute material got approved last quarter and it’s not clear everyone downstream knows about it.

None of these are unreasonable questions. Every one of them is legitimate. The trouble is that answering all five usually means five different people going to five different systems, or worse, going to the one person who happens to remember the answer off the top of their head. And that’s just one event. A mid-sized plant generates thousands of these in a single shift.

This is also why manufacturers who feel like they’ve already solved this problem sometimes haven’t, not fully. Connecting systems so information moves between them quickly is a real accomplishment, and it’s not nothing. Machines that used to require someone to walk over and check a gauge now report status automatically. Material movements get logged the moment they happen instead of at the end of a shift. That’s genuine progress, and it shows up in faster reporting and fewer transcription errors. But speed of movement isn’t the same thing as shared meaning. A message can travel from a machine to an MES to a dashboard in under a second and still leave five different people with five different half-formed pictures of what it actually means for their part of the job.

Why this Keeps Happening Even in Well-Run Plants

None of this is really a knock on any particular team or plant. It’s a structural issue, and it shows up even in operations that are otherwise disciplined and well managed. A plant can have excellent people, solid processes, and genuinely good software in every individual category, engineering, quality, planning, maintenance, and still lose hours every week to exactly this kind of reconciliation work, because good software solving one problem well doesn’t automatically solve the problem of connecting it to everything else.

It also explains why so much digital investment in manufacturing has felt like running in place. Leadership approves a new analytics platform, a new MES module, a new planning tool, each one promising better visibility, and each one genuinely delivers on that promise within its own lane. What rarely gets promised, and rarely gets delivered, is a way for all of those lanes to share the same underlying picture of the operation. So, visibility keeps climbing, department by department, while the actual experience of making a cross-functional decision, the kind that touches engineering, quality, and planning all at once, stays roughly as slow and manual as it’s always been.

What Actually Fixes it

Not another dashboard. Not another integration project either. Getting systems to move data faster is a genuinely different problem from making sure every one of them interprets that data the same way. Manufacturing has made real progress on the first matter. The second is still one of the bigger open opportunities in the industry, and it’s the one that decides whether all that hard-won visibility ever turns into faster, more confident decisions on the floor.

It's worth being specific about what this looks like in practice because it’s easy to move along with the idea and still not change anything. It means defining, once, how a product relates to the process that builds it, how that process relates to the materials and equipment it consumes, how those resources relate to the people qualified to run them, and how all of that relates to the customer commitment sitting behind it. Define those relationships in one place, and every application, every report, every person can draw from the same understanding instead of reconstructing their own version of it every time something happens on the floor.

That’s really where the rethink must start, not as a set of departments passing data back and forth, but as one connected operational system where every event carries enough context for someone to understand what it means. It won’t happen by adding one more report to the pile, and it won’t happen by waiting for the next AI tool to sort it out on its own either. It happens by treating the relationships between products, processes, materials, equipment, and people as something worth building deliberately, instead of something every system reinvents on its own. More on what that looks like in practice in our next blog.