What is data collection, really?
At its core, data collection is the act of gathering information in a structured way so you can answer a question. That sounds obvious — but it's the part most manufacturers skip. They install sensors, hook up PLCs, pipe data into a historian, and then ask: now what?
The question has to come first. Always. Technology follows methodology — and if you don't know what you're trying to answer, no amount of data infrastructure will help you.
Start with a question, not a platform
The biggest mistake we see on the shop floor isn't bad data. It's unfocused data collection. A plant will spend months wiring up a data acquisition system and end up with thousands of data points and zero actionable insight.
Retail figured this out decades ago. When a retailer runs a focus group, they don't ask "tell us everything about your shopping experience." They ask a specific question — and they design the entire data collection effort around getting that question answered cleanly.
Manufacturing needs the same discipline. Before you collect anything, decide on one concise question you want answered. Then work backwards.
What retail does well — and the manufacturing equivalent
Focus groups → Interviewing your assets and IIoT devices
Retail focus groups extract qualitative insight from people. On the shop floor, your machines are the people. Modern IIoT-connected equipment can tell you things no operator can — vibration signatures, temperature drift, cycle time micro-variations. The trick is asking the right questions of the right equipment. A machine interview only works if you know what you're listening for.
Surveys → Selective data collection with operator input
Retailers send targeted surveys — short, specific, to the right customers at the right time. The manufacturing equivalent is selective data capture: not logging everything from every sensor, but designing operator-entry workflows that capture exactly the data points that matter. A well-designed digital quality check form is a survey. A paper traveller with 40 fields nobody reads is noise.
Historical customer data → Linking recipes and shift details
Retailers merge transaction history with customer profiles to find patterns. Manufacturers can do the same thing by linking production data to context — the recipe running, the shift, the operator, the tooling age. Raw machine data alone is rarely actionable. Contextualised data is where the insights live.
Historian data → PLC logs with meaning
A process historian is the manufacturing equivalent of a data warehouse. But just like a retailer's data warehouse needs clean, tagged, well-structured data to be useful, a historian needs context. PLC alarm logs with no timestamps, no part numbers, and no cross-reference to quality outcomes are the industrial version of a spreadsheet nobody opens.
Technology follows methodology. If you don't know what question you're answering, the technology just gives you more places to be confused.
Five steps to better manufacturing data collection
- Decide on a concise question you want answered. Not "how is Line 3 doing?" but "why does Line 3 miss its hourly target on the first hour of the afternoon shift?"
- Be specific about the data you need to answer it. List the exact data points. Resist the urge to collect "while you're at it" data.
- Plan to retrieve data from more than one source. The answer almost always lives at the intersection of machine data, operator input, and production context.
- Invest first in the process, then choose the technology. A well-designed paper-based check sheet beats a poorly designed digital form every time. Get the methodology right, then automate it.
- Build on each success before moving to the next question. Answer one question well. Use that success to justify the next. Incremental, question-driven analytics compounds over time.
Manufacturing data is hard — not because the data doesn't exist, but because the environment is complex, changeable, and unforgiving of sloppy methodology. The plants that get the most out of their data are the ones that treat every analytics initiative like a retail focus group: with a clear question, a defined audience, and a commitment to acting on what they find.