Home  /  Blog  /  Quality Data
Quality Data

your best data is the data you're throwing away

Patrycja Pezan  ·  Aug 17, 2026

Most operations are careful about certain data. Defects get counted, scrap gets logged, returns get recorded. These are the things that went wrong badly enough to force themselves into a system. But the richest information about a process is usually not in any of that. It is in the events that almost went wrong, the adjustments people make to keep things running, the parts that were marginal but usable, the near misses that resolved without a formal record. That data is where the early warning lives, and in most operations it is thrown away, because nothing was built to catch it.

the signal before the failure

Almost no failure arrives without warning. Before the defect there was drift, a process edging toward its limit. Before the escape there was a near miss, a part that barely passed. Before the breakdown there was a machine behaving a little differently that an operator noticed and worked around. These are the early signals, and they carry far more value than the failures they precede, because they are the chance to act before the cost is incurred. Yet they are exactly the data that does not get captured, because near misses, by definition, did not produce a bad outcome, so nothing forced them into a record. The failure gets logged. The warning that preceded it evaporates.

the knowledge inside the workaround

Some of the most valuable discarded data is the compensation your best people do without thinking. An operator feels the process drifting and nudges a setting. Another knows this material runs differently and adjusts. These adjustments are real information about how the process actually behaves, where its weaknesses are, what it takes to keep it in line. And almost none of it is recorded anywhere, because to the person doing it, it is not data, it is just running the job. The process is quietly telling you exactly where it is fragile, through the daily interventions people make to hold it together, and you are throwing that message away every shift because nothing captures it.

why we only keep the failures

The reason is structural. Systems get built to record outcomes that have consequences: the scrap that costs money, the defect that triggers a report, the return that a customer initiated. The near miss has no immediate consequence, so no process was ever designed to capture it, and asking busy people to log things that did not go wrong feels like overhead with no payoff. So the operation optimizes its data collection around failures and blinds itself to everything that precedes them. It ends up with a rich record of problems after they became expensive and almost nothing about the cheap, early moment when they could have been prevented.

capturing the quiet signal

The operations that get ahead of problems find ways to capture the data everyone else discards. They make it easy and worthwhile for an operator to flag a near miss, a part that barely passed, a machine acting strangely, and they treat those flags as valuable rather than as noise. They pay attention to the adjustments their people make and ask why the process needs them. They look at the distribution of their measurements, not just the pass or fail, so they can see drift long before it becomes a defect. None of this requires exotic technology. It requires deciding that the early, quiet, consequence-free signal is worth capturing, which is a decision most operations never make.

the data is already there

The encouraging part is that this data already exists in your operation right now. Your people are noticing the near misses, making the adjustments, seeing the drift. The information is being generated every shift. The only question is whether anything captures it or whether it disappears into the normal flow of the day, remembered by no system and acted on by no one until the warning it contained finally becomes a failure expensive enough to get logged. You do not have to generate better data. You have to stop throwing away the best data you already have.

What is your operation noticing every day that it never records, and what would you see coming if you started keeping the signals instead of only the failures?

Thanks for reading.

I write about quality, manufacturing, and the lessons the floor teaches. If this resonated, follow along on LinkedIn and tell me what it brought up for you.

Connect on LinkedIn →