We’ve walked the floors of a lot of plants. We’ve sat in a lot of conference rooms where a plant manager pulls up a dashboard on a 65-inch screen and says, with genuine pride, “We built this ourselves.” And then we ask one question — “Where did this number come from?” — and the room gets quiet.
That moment tells us everything. The dashboard looks great. The underlying data is a mess. And somewhere upstream, millions of dollars of investment decisions are being made on numbers that nobody can fully defend.
This isn’t a technology failure. It’s a sequencing failure. Manufacturers are investing in the wrong parts of their data stack, in the wrong order, for understandable but costly reasons.
Here’s what we see happening — and what actually works.
The Seduction of the Dashboard
The most common mistake we see is leading with visualization. A leadership team sees a competitor’s operational dashboard at a conference, or a vendor demo shows real-time OEE by line, and the reaction is immediate: we need that.
So they buy it. Or they build it. And within six months, they have a beautiful interface sitting on top of data that’s incomplete, inconsistently defined, and manually entered by shift supervisors who have seventeen other things to do.
The dashboard wasn’t the problem. The sequence was.
Visualization is the last mile of your data stack, not the foundation. When you build it first, you spend most of your budget on something that can only be as good as what feeds it. And what feeds it — the data collection layer, the integration layer, the governance layer — gets underfunded because you’ve already spent the money on the thing the executives can see.
The Real Layers, In Order
When we work with a new manufacturing client, we think about the data stack in four layers. Most organizations have inconsistent investment across all four, but the damage compounds when you skip the lower ones.
Layer 1: Data Collection and Source Integrity
This is where the data is born. PLCs, SCADA systems, MES platforms, ERP transactions, manual entry at the line. If the data is wrong here — if a sensor is miscalibrated, if operators are rounding to the nearest five, if your MES timestamps are 11 minutes off from your ERP — nothing downstream can fix it. You can clean it, flag it, and work around it, but you cannot manufacture truth out of bad source data.
Most manufacturers underinvest here because it’s unglamorous. Fixing a sensor calibration process or standardizing how downtime codes get entered doesn’t show up in a board presentation. But it is, without question, the highest-leverage place to spend your first dollar.
Layer 2: Integration and Data Movement
Manufacturing environments are notoriously fragmented. You might have one ERP, two MES systems from different acquisitions, a quality system that nobody has touched since 2014, and three different ways of defining “shift.” Getting data from these systems into a common environment — reliably, on a schedule you trust, with clear lineage — is hard, expensive, and invisible to most stakeholders.
This is where we see manufacturers make their second most common mistake: assuming their ERP vendor’s integration tools are sufficient, or that a business intelligence platform’s built-in connectors will handle the complexity of shop floor data. They rarely do, not at the volume and variety that real manufacturing environments produce.
Layer 3: Data Modeling and Semantic Consistency
This layer is almost always skipped. It’s the layer where you answer the question: what does “yield” mean in this plant, and is that the same definition we use in the other plant?
Semantic consistency sounds academic until you’re trying to consolidate reporting across six facilities and discover that each one calculates OEE differently, two plants define a “production run” in incompatible ways, and nobody wrote any of this down. Now your corporate analytics team is spending 40% of their time reconciling definitions instead of generating insight.
A data model — even a simple one, even a documented spreadsheet — is worth more than most manufacturers realize until they don’t have one.
Layer 4: Reporting, Analytics, and Visualization
This is where most manufacturers want to start. This is where the vendors want to sell you. And it’s legitimate — this layer is where analysts do their work, where operators see their metrics, where executives make decisions. But it is entirely dependent on the three layers beneath it.
When the lower layers are solid, this layer is fast to build, easy to maintain, and actually trusted by the people who use it. When they’re not, you end up with the quiet room we described at the beginning of this piece.
Why Manufacturers Sequence This Wrong
It’s worth being clear that this isn’t a case of poor judgment. There are structural reasons manufacturers invest in the wrong order.
Vendor incentives point up the stack. The vendors with the largest sales teams and the most compelling demos are selling Layer 4 products. Dashboard and BI vendors are sophisticated at showing executives what’s possible. The infrastructure and integration vendors who operate in Layers 1 and 2 tend to be more technical, harder to evaluate, and less adept at selling to the C-suite.
ROI is easiest to articulate at the top. “A real-time OEE dashboard will help our supervisors make faster decisions” is a sentence any plant manager can write in a capital request. “A reliable, well-integrated data pipeline with consistent semantic definitions across facilities” is harder to quantify, even though the latter is what makes the former worth anything.
Quick wins create political momentum. A dashboard ships in eight weeks. An integration project takes eight months. In a manufacturing environment where the CFO wants to see results before the next budget cycle, there’s real pressure to show something visible, something now. That pressure is understandable. It’s also how you end up with dashboards that operators don’t trust and executives don’t use.
What a Better Sequence Looks Like
We’re not suggesting manufacturers should spend three years getting their data infrastructure perfect before building a single report. That’s its own kind of mistake — analysis paralysis in infrastructure form. But there is a smarter path.
Start narrow, not shallow. Pick one process, one line, one plant. Go deep. Get the data collection right. Build the integration properly. Define the terms. Then build the dashboard. Prove the model works end-to-end on a small scope before you try to scale it. This approach produces something you can actually trust, and something you can replicate — which turns out to be far more valuable than a beautiful enterprise dashboard that no one believes.
Instrument the data before you analyze it. Before you ask “what does our OEE look like?”, ask “how confident are we in how we’re measuring downtime?” That second question will save you from the first question’s answers being wrong.
Make data governance boring and operational. Governance sounds like a committee. In manufacturing, it should be as routine as a PM schedule. Who owns each data source? What does each KPI mean, exactly? Who approves changes to definitions? This doesn’t require a data governance framework document. It requires someone who is accountable and a simple record of decisions made. Start there.
Buy for the layer you’re in, not the layer you want. If you don’t have reliable data collection and integration, the most important vendor conversation you can have is with someone who solves that problem — not with someone who sells visualization. The visualization vendor will still be there when you’re ready.
The Question Worth Asking Before Every Data Investment
Before approving any data project — tool, platform, dashboard, or initiative — we ask our clients one question:
If this project succeeds, what will we be able to do that we can’t do today, and what has to be true about our existing data for that to matter?
That second clause is the one that matters. It surfaces the dependencies. It reveals whether you’re building on a foundation or papering over one.
Manufacturers who get their data stack right aren’t the ones who bought the best tools. They’re the ones who invested in the right sequence — who built trust in their data before they built displays of it. That trust, once established, compounds. Every analytics initiative is faster, cheaper, and more impactful when the layers beneath it are solid.
The plants we’ve seen do this well all share one trait: someone, at some point, was willing to delay the dashboard and fix the data first. It’s never the popular decision. It’s always the right one.
Get Started on Building Your Data Stack
Lasso works with mid-market and enterprise manufacturers on data strategy, analytics infrastructure, and operational intelligence. If you’d like to talk through where your data stack investment is sequenced, get in touch.
