Manufacturing KPIs Every Pump Plant Owner Should Track

The six that matter

Six is deliberate. A plant that genuinely manages these six outperforms one that reports forty and acts on none.

OEE, and when a simpler measure is better

Overall Equipment Effectiveness combines availability, performance and quality into one figure. It is useful on a genuine bottleneck machine and misleading almost everywhere else, because improving OEE on a non-constraint produces inventory rather than output.

For most pump plants a more honest daily measure is output against plan by model, with downtime reasons recorded. That tells you whether you built what you intended and why not, which is the actual question. Reserve full OEE for the one or two machines that genuinely constrain the plant.

Whichever you use, record downtime reasons as categories rather than free text, or the data will not aggregate into anything actionable.

On-time delivery, defined honestly

This is the measure most often quietly corrupted. Common distortions:

• Measuring against a revised date. If the date can be moved when it is about to be missed, the metric approaches one hundred percent and means nothing.
• Measuring dispatch rather than receipt. Reasonable if your responsibility ends at the gate, but be explicit.
• Counting partial shipments as on time. A part shipment is usually a miss from the dealer's perspective.

Define it as delivery complete against the first promised date, then live with the uncomfortable number that produces. That figure is the one that correlates with dealer satisfaction, and it is the one worth improving. Our production planning article covers what usually drives it.

First pass yield and where scrap hides

First pass yield is the proportion of units that pass all checks without rework. It is more useful than a rejection percentage because it captures rework, which is often the larger cost and is frequently invisible.

Measure it at the stages that matter for a pump: incoming casting acceptance, post-machining, and final test. Reporting each separately shows where the loss originates, which a single plant-level figure cannot.

The common trap is that rework performed informally at the station never gets recorded, so yield looks excellent while the plant is quietly absorbing a large amount of correction. If your first pass yield is very high and your delivery still slips, that is usually what is happening.

Inventory turns and model margin

Inventory turns tell you whether cash is moving. Track overall and by class, because a healthy aggregate can conceal a large pile of slow-moving castings offset by fast-moving finished goods. Watch non-moving stock separately and make an explicit decision on it each quarter.

Model margin is the number most likely to change commercial behaviour. Revenue less rolled-up material, machining, job work and overhead cost, by model and HP variant. Manufacturers computing this properly for the first time frequently discover that certain variants they actively promote lose money, usually because material cost has drifted while the price list has not.

Both depend on costing that reflects reality, including landed cost and job work accrual, which our procurement article discusses.

Making dashboards actually get used

Dashboards fail for predictable reasons. They are built for the person who requested them rather than the person who must act. They show too much. They update too slowly to be trusted. Or nobody has a standing appointment to look at them.

What works:

• Role-based views. The plant head, the sales head and the owner need different screens, not the same one filtered.
• Few numbers, clearly defined. Every metric should have a written definition that survives a disagreement.
• Live from transactions, so nobody maintains a parallel spreadsheet.
• A standing review rhythm. Fifteen minutes daily on output, thirty minutes weekly on delivery and quality, an hour monthly on margin and cash.

The rhythm matters more than the tooling. A modest dashboard reviewed consistently beats an elaborate one nobody opens. See analytics and dashboards for how these are normally structured.

Conclusion

Measurement in a pump plant fails in two directions: measuring nothing, or measuring so much that nothing gets acted upon. Six numbers, defined honestly and reviewed on a fixed rhythm, will tell an owner most of what they need. Are we building what we planned, are we keeping our promises to dealers, are we building it right first time, is cash moving, which models actually earn money, and are we collecting what we billed. The discipline that makes them work is definitional honesty, particularly on on-time delivery, and a standing appointment to look at them. Retire any measure that has not changed a decision in six months. It is costing you attention that the remaining ones deserve.