“Margin dropped three points this quarter.” It’s an important number, but it doesn’t tell you what to do. Was it raw material prices, scrap, overtime, a machine that sat idle longer, or a customer demanding rush deliveries? With that single figure, the leadership meeting turns into opinions, and every department defends its own.
That’s exactly what operational KPIs are for: tracing a result back to its cause. In this article we explain which metrics to track in an operation, how to organize them so they point to where the problem is, and the mistakes that turn a dashboard into a report nobody uses.
Result metrics and process metrics
Before choosing what to measure, separate two types of metrics:
- Lagging (result) metrics: they tell you what happened. Margin, unit cost, sales, customer complaints. They arrive late and don’t tell you why.
- Leading (process) metrics: they tell you what’s happening in operations and anticipate the result. Downtime, defects per shift, schedule attainment, changeover time.
A good operational dashboard has a few result metrics at the top and, under each one, the process metrics that explain it. If you only have the first kind, you find out about the problem after it has cost you. If you only have the second, you measure a lot without knowing what matters.
The operational KPIs that actually matter
You don’t need forty metrics. For most operations, one or two in each of these five dimensions is enough:
| Dimension | KPI | How to calculate it | What it tells you |
|---|---|---|---|
| Quality | First pass yield (FPY) | Good units without rework ÷ Units produced | How much comes out right without fixing anything. |
| Quality | Complaints or returns | Complaints ÷ Orders delivered | What the customer actually sees. |
| Delivery | On time in full (OTIF) | Orders on time and complete ÷ Total orders | Whether you deliver what you promised. |
| Delivery | Lead time | Days from order to delivery | How fast you respond. |
| Productivity | OEE | Availability × Performance × Quality | How much equipment capacity turns into good product. |
| Productivity | Units per labor hour | Good units ÷ Hours worked | Labor productivity. |
| Cost | Unit cost | Production cost ÷ Good units | The effect of everything above, in money. |
| Cost | Overtime | Overtime hours ÷ Regular hours | Often a symptom of another problem. |
| People and safety | Absenteeism and incidents | Absences ÷ Scheduled days; incidents per 200,000 hours | The health of the team running the operation. |
If your operation carries inventory, add turnover, days of inventory, and fill rate. We explain them in 7 signs your inventory is eating into your margin.
In service companies the logic is the same: quality (cases resolved the first time), delivery (service-time compliance), productivity (cases per person per day), and cost (cost per case or per customer served).
OEE: the metric that says the most about a plant
OEE (Overall Equipment Effectiveness) deserves its own section, because it’s one of the few metrics that tells you what kind of loss the problem is. According to OEE.com, it’s calculated by multiplying three factors:
- Availability: time the equipment actually ran ÷ planned production time. Captures stops, breakdowns, and changeovers.
- Performance: actual output versus what would have been produced at ideal speed. Captures small stops and slow running.
- Quality: good parts ÷ parts produced. Captures scrap and rework.
Example: a line has 480 planned minutes and is stopped for 60 of them (87.5% availability). While running, it works at 90% of its ideal speed (90% performance), and 97% of what it produces is good (97% quality).
OEE = 87.5% × 90% × 97% ≈ 76%
The value isn’t the 76%, but seeing that the biggest loss is availability. The conversation is no longer “the line doesn’t produce enough”; it’s why it sits idle for an hour every shift.
What’s a good OEE? The same source notes that 85% is known as “world class” in discrete manufacturing (based on 90% availability, 95% performance, and 99% quality), but that most plants are closer to 60%. Rather than chasing the benchmark, compare yourself against your own history.
How to use KPIs to find where the problem is
1. Build a metrics tree
Start with a result (for example, unit cost) and ask: what does it depend on? Materials cost, labor, scrap, and equipment productivity. Each of those has its own metric. When cost goes up, you know which branch to look in.
2. Break it down before drawing conclusions
Averages hide problems. Review each metric by line, shift, product, customer, or supplier. It’s very common for 80% of defects to come from two products, or for one shift to have twice the downtime of the others.
3. Rank the causes with a Pareto chart
Record the causes of each loss (stops, defects, late deliveries) and sort them from largest to smallest. Tackling the top two or three usually solves most of the problem.
4. Go and see
Data tells you where to look; you find the cause where the work happens, observing the process with the team that runs it. It’s the gemba principle we explain in the 8 wastes of Lean: almost every loss a KPI measures is one of those wastes.
Rules for a dashboard people actually use
- Few metrics per department. Four to six ongoing KPIs per department is enough. If you need to move one significantly, turn it into a change goal, as we explain in our guide to OKRs vs. KPIs.
- A written definition for each one. Formula, data source, frequency, and owner. If two departments calculate OEE differently, the discussion will be about the number, not the problem.
- Fresh data, close to operations. A metric that arrives a month late only helps you explain. On the floor, visual boards by shift let you react the same day.
- A target and a baseline. Without them, you don’t know whether 92% is good or bad.
- A review that ends in agreements. A short daily huddle on the floor, a weekly review per department, and a monthly review with leadership. Every significant deviation should leave with an owner and a date.
Common mistakes when measuring operations
- Measuring what’s easy instead of what matters. Units produced are easy to count; good units delivered on time, less so, but those are the ones that matter.
- Metrics that contradict each other. Rewarding a supervisor for volume and not quality produces volume with scrap.
- Dashboards without owners. If nobody is accountable for a number, nobody moves it.
- Confusing the symptom with the cause. High overtime is almost never the problem; it’s the consequence of downtime, absenteeism, or poor planning.
Frequently asked questions about operational KPIs
How many KPIs should a plant have?
At the leadership level, five to eight result metrics. In each department, four to six process metrics. More than that usually means nobody reviews them all.
Does OEE work if my processes are manual?
It works with adjustments. In manual processes, labor productivity and schedule attainment are usually more useful, although the logic of separating losses into stops, speed, and quality still applies.
Do I need software to track my KPIs?
Not to get started. A well-defined spreadsheet and a visual board on the floor deliver very good results. Software makes sense once you know what to measure and need to automate data capture.
How often should I review the metrics?
Process metrics, daily or by shift. Result metrics, weekly or monthly. What matters is that the review frequency lets you act before the problem reaches the customer.
Conclusion
An operational KPI isn’t there to tell you how well you’re doing, but where to act. A few metrics, organized from result to cause, broken down, and reviewed with discipline, turn the leadership meeting into a conversation about concrete problems instead of opinions.
At Total Consulting we help companies design operational dashboards and use them to improve productivity, quality, and delivery through our Operational Excellence service. If your numbers tell you something is wrong but not where, schedule a free diagnostic session.
