Manufacturing KPIs That Actually Drive Improvement

Not just definitions — which metrics matter, which are vanity, and how to implement them correctly

The Problem with Manufacturing KPIs

Walk into any manufacturing plant and ask about their KPIs. You’ll likely hear about OEE, maybe some quality metrics, and perhaps on-time delivery. But here’s the uncomfortable truth: most manufacturers are measuring the wrong things, or measuring the right things the wrong way.

I’ve seen plants with beautiful dashboards showing OEE at 85% while they’re hemorrhaging money. I’ve seen others obsessing over metrics that look good in board meetings but drive zero improvement on the shop floor.

The difference between successful manufacturers and the rest isn’t which KPIs they track — it’s how they implement and act on them.

The KPIs That Actually Matter

Let’s cut through the noise. Here are the metrics that drive real improvement when implemented correctly:

1. OEE (Overall Equipment Effectiveness)

What it is: A standardised efficiency calculation that combines three factors:

Availability — Is the equipment running when it should be?
Performance — Is it running at the right speed?
Quality — Are we making good parts?

Why it matters: OEE abstracts away the differences between equipment types. You can take OEE from individual machines, roll them up into work centres, then areas, then sites, all the way to the boardroom. It’s the most generic metric that works for anything — discrete, batch, or continuous processes.

The catch: OEE is an *operational* metric, not a *strategy* metric. It tells you how effectively equipment is running, but it won’t tell you if your smart factory strategy is working. I’ve seen people use OEE to justify digital transformation investments — that’s the wrong application entirely.

Implementation tip: The biggest mistake? Only tracking major downtime events. Nobody tracks short downtime on paper, and even if they do, operators are pencil-whipping it. When you start collecting downtime data digitally, you get actual distributions of mean time to repair (MTTR) and mean time between failure (MTBF). That’s where the insights live.

2. First Pass Yield (FPY) / Quality Rate

What it is: Good parts divided by total parts. Simple.

Why it matters: Quality is often the easiest of the three OEE components to calculate — as long as you’re actually tracking total parts in and good parts out of each production line.

The catch: Most quality systems aren’t getting any data from the plant floor. They’re disconnected from what’s actually happening at the edge. The biggest gap in quality management is the lack of real-time information flow from machines to quality systems.

Different calculations for different processes:

Discrete manufacturing: Good parts ÷ Total parts
Batch processing: Actual yield ÷ Target yield (e.g., 990 litres out of 1000 = 99% quality)
Continuous processes: Volume-based calculations where cycle time approaches zero

Real-world insight: One client was off by 20% on their yield calculations until they deployed ML-powered camera systems to measure output. No new sensors needed — just existing cameras trained to count and classify product as it came off the line.

3. MTBF and MTTR (Mean Time Between Failure / Mean Time To Repair)

What they are:

MTBF — When running, how long before the next failure?
MTTR — When something breaks, how long to fix it?

Why they matter: These metrics drive maintenance and engineering decisions. Which equipment should we invest in to improve? Where are the reliability problems?

The multi-layer problem: You need to track these at multiple levels:

Machine layer — Raw equipment data
Human layer — When was maintenance notified? When did someone show up? How long to repair? When did production resume?

This is where maintenance and engineering teams get nervous. If they’re making capital decisions based on these numbers, they need to trust the data. And most don’t, because they know the data is incomplete.

The hidden value: When you calculate MTBF and MTTR properly, you can start doing predictive maintenance. You don’t even need to make a business case first — just start collecting the data, and patterns emerge. Machine learning can’t analyse what you’re not collecting.

4. On-Time Delivery (OTD)

What it is: Percentage of orders delivered when promised.

Why it matters: This is the metric your customers care about. Everything else is internal optimisation.

The connection: OTD failures are usually traced back to production issues. Late shipment? Check the timeline — you’ll find downtime, quality rejects, or scheduling problems upstream.

5. Downtime Classification

What it is: Categorised reasons for every production stoppage.

Why it matters: You can’t improve what you can’t classify. Knowing *that* you had downtime is useless without knowing *why*.

Implementation approaches:

Manual classification: Operators select reasons via the interface when machines stop
Automatic classification: Rules-based triggers from alarms and conditions
Hybrid: Automatic for known patterns, manual for exceptions

Pro tip: Machine Metrics and similar systems can automatically classify setup time, material unavailable, and other standard categories — giving you better reporting without operator burden.

The Vanity Metrics Trap

Not all KPIs are created equal. Some look great on dashboards but drive no improvement:

Dashboard metrics without action triggers — If seeing a number doesn’t prompt someone to do something, why are you measuring it?
Lagging indicators only — Knowing last month’s OEE is useless if you can’t see today’s performance in real time.
Metrics without context — 85% OEE sounds good until you realise the benchmark for your industry is 92%.
ROI for digital transformation — Classic ROI is often the wrong metric for Industry 4.0 initiatives. You may not see traditional ROI for 2+ years while building foundational capabilities.

Different KPIs for Different Stages

Here’s what most people miss: success is measured differently at different stages of your journey.

When you’re climbing the first mountain (digitising operations), you use different KPIs than when you’re measuring digital supply chain maturity. The measurement criteria for your initiative should evolve as your capabilities mature.

Early stage:

Data collection coverage
System connectivity
Operator adoption

Mid stage:

OEE improvement
Downtime reduction
Quality improvement

Advanced stage:

Predictive maintenance effectiveness
Supply chain synchronisation
Revenue per employee

How to Get Started

1. Pick ONE metric to implement properly — OEE is usually the best starting point because it cascades into availability, performance, and quality insights.

2. Digitise the data collection — Paper-based tracking doesn’t work. You need real-time, automatic capture where possible.

3. Close the loop — Every metric should trigger an action. If availability drops below the threshold, who gets notified? What happens next?

4. Build trust in the data — The number one objection from maintenance and engineering: “How do we know we can trust this?” Address data quality before pushing for decisions.

5. Iterate your KPIs — As you collect more data, new KPIs become possible. Football clubs transformed player recruitment once they had years of detailed match analytics — tracking everything from distance covered to pass completion rates. Your manufacturing metrics should evolve the same way.

The Bottom Line

The manufacturers who win aren’t tracking more metrics — they’re tracking the right metrics, implementing them correctly, and actually acting on what they see.

Stop chasing dashboard vanity. Start with one KPI, implement it end-to-end, and build from there.

Want help identifying which KPIs matter for your operation? Contact us for a free assessment of your current metrics and where to focus first.

Related reading: