Team Scores Keep Rising—Why Are the Real Problems Harder to See?

Why Are the Real Problems Harder to See When Team Scores Keep Rising?

At the monthly meeting, several teams' composite scores exceeded 90 points, with production, yield, equipment, and safety all showing green indicators. However, delivery times remain volatile, rework is still increasing, and fewer people are actively reporting anomalies on the shop floor.

When an overall score simultaneously affects ranking, bonuses, and accountability, it becomes more than just a measurement tool. Teams will naturally focus on metrics that are easier to improve and explain, while leaving difficult issues outside the statistical scope.

Performance indicators are not better when they are more concentrated. An effective performance system should allow managers to see trade-offs between objectives and enable the shop floor to know when problems must be exposed.

Total Scores Reduce Risks of Different Natures into One Number

Can a two-point increase in production offset a critical quality escape? Can an increase in equipment utilization offset insufficient maintenance? These indicators differ in units and impact cycles, and simple weighting can easily create false comparability.

NIST's research on smart manufacturing performance includes agility, asset utilization, and sustainability in separate classification frameworks, showing that manufacturing systems are inherently multi-objective and cannot be fully represented by a single outcome.

Another NIST study points out that productivity, quality, and maintenance KPIs are inherently related. A change in one indicator could represent improvement or simply shifting the cost to another area.

The More Beautiful the Color, the More We Need to Question the Denominator and Boundaries

Improvement in yield may be due to rework items being removed from statistics; reduction in downtime may be due to short stops not being reported; increased delivery fulfillment may be due to repeatedly modified commitment dates.

Managers should not first ask whether numbers have met targets, but rather whether the subject, time, status, and exclusions are consistent. Without a stable metric definition, rankings are merely comparing different versions of facts.

The NIST Operational Performance Project emphasizes that one should first define system boundaries and references before using operational data to analyze problems. In other words, metrics must serve specific decisions, not the other way around—first having a dashboard and then searching for data to fill it.

Let Exceptions Be Input, Not a Reason for Deducting Points

If reporting downtime reduces a team's score, the most rational individual choice is to delay reporting. If identifying quality risks only increases rework responsibility, employees will wait for the inspector to initiate action.

This is not simply an attitude issue, but rather a misalignment between incentives and responsibilities. Recording timely detection, proper isolation, and complete handover separately can distinguish between 'exposing problems' and 'causing problems.'

Replace a Ranking List with Three Layers of Indicators

The first level looks at the bottom line: safety, regulations, and critical quality. Failure in any of these cannot be offset by other scores. The second level looks at results: delivery dates, costs, and output. The third level looks at processes: abnormal response, closure time, and recurrence rate.

Teams can compare common metrics, but improvement areas should consider equipment, product, and task structure. Comparing high-mix low-volume teams with stable batch teams using the same cycle time target for ranking may mislead decision-making.

Factory owners should select a high-scoring team each month and ask three reverse questions: What anomalies were not included in the score, what costs were transferred to the next process, and what good results cannot be consistently replicated.

Turn these three questions into a retrospective checklist, verifying metric fields, exclusions, and responsibility boundaries. Abnormalities should be escalated by team leaders, handled by specialized roles, and confirmed closed by the plant manager. They should not be passed on without responsibility.

Not all factories need complex models. Small factories can start by removing a scoring item that encourages concealment. Growing factories should clearly define data sources, while mature factories should also check the relationships between metrics.

Responsibilities and authorities should align. If reporters are not accountable or handlers lack authority, frontline staff will avoid responsibility. Separating escalation, handling, and closure in records can reveal where management actions are blocked.

A truly healthy dashboard is not always green; it is healthy when red lights are pressed, taken over, and closed by someone. Total scores can provide an overview, but they must never replace a manager's judgment of shop-floor relationships.

I am Huang Chaoxing, inventor of the TreeMES stack-light MES system, and I focus on analyzing shop-floor management and digitalization trends in manufacturing.