Manual Pareto vs real-time analysis: finding the true cause

The Pareto, a classic of continuous improvement

The Pareto chart is one of the most widely taught tools in continuous improvement, and for good reason. It rests on a robust principle: a minority of causes explains the majority of effects. Applied to production losses, it consists of ranking losses by category and by importance, so that effort concentrates on the few causes that weigh the most. It’s simple, visual and effective to get started. (OEE, Overall Equipment Effectiveness, is the English name for what French manufacturers call TRS.)

This tool has proven its worth and remains valuable. It prevents scattering, ranks the projects to tackle and gives a clear logic of action. But like any tool, it has limits, and these owe less to the Pareto principle itself than to the way you feed it. A Pareto built by hand, from predefined categories and manual readings, inherits all the limits of that raw material.

What the Pareto does well

The manual Pareto ranks known, categorised losses effectively. If you already know how to sort your stoppages into broad families – breakdowns, changeovers, quality defects – it clearly tells you which to treat first. For an initial diagnosis, it usefully orients the projects and gives a direction of action the teams can share.

That strength is also its boundary. The Pareto only sees what you managed to categorise in advance. It files losses into boxes defined by the analyst, and it’s only as good as those boxes. If an important loss doesn’t fit any planned category, or if it results from the combination of several factors, the manual ranking misses it. The tool ranks the known well, but it’s blind to the unexpected.

What the manual Pareto doesn’t see

The main limit of the manual Pareto is that it reasons one category at a time, whereas the costliest losses are often born from the interaction of several parameters. A defect that only appears with a certain material batch, in a certain time window, and with a certain setting: that’s a multi-parameter cause no ranking by isolated category can ever reveal.

These combined causes escape, by their very nature, a Pareto built by hand. You can sort stoppages by type all you like, you’ll never see that it’s their conjunction with other factors that explains the problem. Yet it’s precisely in those interactions that the most stubborn losses often hide – the ones that resist conventional projects because their true cause was never identified, for lack of a tool able to see it. A symptom you can name is not the same as a cause you can act on, and the manual Pareto stops at the first.

Why manual logging makes the limit worse

The weakness of the manual Pareto is compounded by that of its data source. When losses are logged by hand, at the end of the shift, a large share of them isn’t even recorded: micro-stops and under-pace, too short to be noted, never appear in the ranking. The Pareto is then built on an already truncated view of reality.

You end up with a chart that ranks the visible losses correctly but ignores a whole part of the problem. The decisions that follow concentrate on the tip of the iceberg, while the invisible losses keep running. Improving the Pareto therefore first means improving its data: moving from a partial reading to a complete, continuous measurement. Otherwise you keep refining the ranking of a reality you’ve only half captured, and the conclusions inherit that blind spot whatever care you put into the chart itself.

Moving to continuous analysis

Real-time measurement changes the game. By recording every event to the second, with its timestamp and its context, it lets you tie losses to precise situations, far beyond what a manual Pareto could ever file. You’re no longer content to know that a type of loss is frequent: you can explore when, under what conditions and in connection with which factors it occurs. Each loss carries its own story, and that story is now available rather than lost at the end of the shift.

This wealth of data opens the way to multi-parameter cause analysis. By cross-referencing losses with the material batch, the time window, the team or the combination of settings, you surface correlations the human eye and the manual ranking simply could not grasp. The Pareto isn’t replaced: it’s fed by infinitely more complete data, which pushes back its limits.

The Pareto isn’t dead, it evolves

It would be wrong to conclude that the Pareto is outdated. The principle stays valid and valuable: concentrating effort on the dominant causes is always the right strategy. What changes is the quality and depth of the data it rests on. A Pareto fed by complete real-time measurement is far more powerful than a Pareto built on partial manual readings.

The evolution therefore points towards an enriched Pareto, able to integrate the invisible losses and to explore the interactions. You keep the prioritisation logic that makes its strength, and you add the completeness and the fineness that only automatic measurement provides. It’s that combination that leads to the true cause, where the manual Pareto stopped at the dominant symptom. The chart looks the same on the wall; what differs is the trust you can place in every bar of it.

From measurement to cause, step by step

The complete approach unfolds like this: measure first, continuously and completely, to see every loss; rank next, following the Pareto logic, to target the dominant ones; then analyse the interactions, to trace back to the real root cause rather than the symptom. Each step rests on the previous one, and all of them rely on the quality of the initial measurement.

It’s that rigour that separates a lasting fix from a temporary patch. As long as you act on the symptom, the problem comes back; when you reach the root cause, it disappears. Hutchinson improved its OEE from 42% to 75% with the same headcount and machines, sensor installed in under an hour. Measurement isn’t an analytical gadget: it’s the foundation on which any serious search for cause can rest, starting with a Pareto finally fed by reality.

One last advantage of this approach deserves a mention: the traceability of the reasoning. When each Pareto priority follows from measured data, and each identified cause rests on objective correlations, the improvement effort becomes transmissible and defensible. You can explain why you chose to treat one loss rather than another, show the gain obtained, and reproduce the method on other lines. A Pareto enriched by measurement isn’t only more accurate: it makes continuous improvement more rigorous and more shareable.

Key takeaways

The manual Pareto usefully ranks known losses, but it doesn’t see multi-parameter causes or the short losses absent from manual readings – precisely where the costliest losses often hide. It isn’t outdated: it must rest on a real, complete measurement. Real-time measurement, by tying losses to their precise context, enriches the Pareto and opens the way to the true root cause.

Measure completely, rank, then analyse the interactions. Hutchinson improved its OEE from 42% to 75% with the same headcount and machines, sensor installed in under an hour. That’s the order that turns a dominant symptom into a cause you can actually eliminate.

FAQ

Is the Pareto outdated for analysing losses?
No. Its principle – concentrating effort on the dominant causes – stays valid and valuable. But it must rest on a real, complete measurement, not on partial manual readings, to deliver its full power.

Why does the manual Pareto miss causes?
Because it reasons one category at a time and doesn’t see the interactions of several parameters (material, time, setting). Yet the costliest losses are often born from those combinations, which a manual ranking cannot reveal.

How does manual logging make the problem worse?
Because a large share of losses – micro-stops, under-pace – isn’t even recorded. The Pareto is then built on a truncated view, and the decisions ignore a whole part of the problem.

How do I find the true root cause?
By starting from a complete real-time measurement, tied to the production context (batch, time, team, settings), to surface the multi-parameter correlations. You rank with the Pareto, then analyse the interactions.

Does measurement replace the Pareto?
No, it feeds it. You keep the prioritisation logic that makes the Pareto’s strength, and you add the completeness and fineness of automatic measurement. It’s the combination that leads to the cause, not to the symptom alone.