Manufacturing Process Efficiency: How to Measure and Improve It

Key takeaways

• Manufacturing process efficiency describes how much of the time, material and capacity in a process actually turns into good product.
• Measure it at three levels: each process step, the bottleneck, and the whole flow from raw material to finished goods.
• In the worked example below, a part needs 14 minutes of work but spends 934 minutes in the process. That is a process cycle efficiency of 1.5%.
• The same example shows why capacity on paper is not capacity in practice. At 80% OEE, the bottleneck makes 72 good units in a shift against a demand of 80.
• Improve in this order: see the flow, protect the bottleneck, cut waiting, then make it right the first time.

What is manufacturing process efficiency?

Manufacturing process efficiency is how well a sequence of process steps converts time, material and equipment capacity into good finished product.

It is a family of measures rather than one number. Each measure answers a different question about the same process.

A process can have fast machines and still be inefficient. Parts can wait for hours between steps, get reworked twice, or pile up in front of one slow station.

Process efficiency vs production efficiency

Production efficiency compares actual output with standard output, usually for one machine or line.

Process efficiency looks along the whole flow: the steps, the waiting between them, the yield at each step and the constraint that sets the pace.

Three levels to measure

Many plants track the first level closely and rarely measure the third. The third level is often where the surprises are.

The key process efficiency metrics

Our guide to manufacturing cycle efficiency goes deeper into the value-added ratio. For the differences between the time measures, see cycle time vs takt time vs lead time.

Worked example: a four-step process

A plant makes a welded, painted assembly in four steps, with one station per step. The numbers are round to keep the math easy, so replace them with your own.

Transport between the steps adds 20 minutes in total. The plant runs a 450-minute shift, and customer demand is 80 units per shift.

Step 1: process cycle efficiency = 1.5%

Value-added time is the time spent actually changing the part. Here we count each full cycle time as value-added, so 2 + 4 + 3 + 5 = 14 minutes.

Lead time adds the waiting and moving: 14 + 60 + 240 + 480 + 120 + 20 = 934 minutes, or about 15 hours and 34 minutes.

PCE = 14 ÷ 934 × 100 = 1.5%. For 98.5% of its time in the process, the part is waiting or moving.

Step 2: rolled throughput yield = 89.4%

RTY = 98% × 95% × 97% × 99% = 89.4%.

Out of every 100 units started, only about 89 pass all four steps the first time. The other 11 or so need rework or become scrap somewhere along the way.

Notice that every step looks good on its own, and the average step yield is about 97%. The losses only become visible when you multiply them, which is why averaging step yields hides the problem.

Step 3: takt time vs the bottleneck

Takt time = 450 minutes ÷ 80 units = 5.625 minutes per unit.

Assembly has the longest cycle time, 5 minutes per unit, so it is the bottleneck on paper. It can make 450 ÷ 5 = 90 units per shift, so capacity looks fine.

Step 4: the bottleneck at real OEE

Now add reality. The assembly station has an OEE of 80%, because of stops, slow cycles and a few rejects.

Fully productive time is 450 × 80% = 360 minutes. At 5 minutes per unit, that is 72 good units per shift, eight short of demand.

To make 80 good units, the bottleneck needs an OEE of 80 × 5 ÷ 450 = 88.9%. A faster cutting machine would not add a single good unit, because cutting is not the step that limits output.

This assumes the other three stations run at 80% OEE or better, so assembly stays the constraint. Welding, the next slowest step, would only take over below 64% OEE.

If the earlier steps keep releasing 80 units a shift, eight of them pile up in front of assembly every shift.

What the example tells you

After three changes

Suppose the plant paints in smaller batches, which halves the wait before paint to 240 minutes. It also releases work to welding in smaller lots, which halves the queue before welding to 120 minutes.

Lead time falls to 14 + 60 + 120 + 240 + 120 + 20 = 574 minutes. That is 360 minutes, or 38.5%, shorter, and PCE rises to 14 ÷ 574 × 100 = 2.4%.

If welding also improves its first pass yield from 95% to 98%, RTY rises to 98% × 98% × 97% × 99% = 92.2%.

None of these changes made a single machine faster. They removed waiting and rework.

Output is still 72 units per shift, because none of these changes touched assembly. That gap only closes when assembly OEE reaches 88.9%.

How to measure process efficiency in five steps

1. Map the flow. Draw every step, queue and move from raw material to shipping in a value stream map, and use a spaghetti diagram for the travel.
2. Time the steps and the waits. Record cycle time at each step and how long parts wait in front of it, tagging a few parts to follow if you have no system data.
3. Record first pass yield per step. Count units that needed rework as failures, even if they were saved later.
4. Find the bottleneck and measure its OEE. Use machine data, not a clipboard, because short stops at the constraint rarely get written down.
5. Calculate the flow measures. Work out lead time, PCE, RTY and takt time, and write down the date and conditions so you can compare later.

10 ways to improve manufacturing process efficiency

The order matters. Fix the constraint and the waiting first, because speeding up a step that is not the bottleneck only builds a bigger queue.

1. Find the real bottleneck

A big queue is a clue, but it can also come from batching, like the paint queue in the example.

Confirm the constraint by comparing each step's real capacity, available time × OEE ÷ cycle time, with demand.

Our bottleneck analysis guide shows the method.

2. Raise OEE at the bottleneck

Every minute the constraint stands still is a minute of output lost for the whole process.

Capture every stop automatically, give each one a reason, and attack the biggest losses first.

3. Protect the bottleneck from starvation

Keep a small, controlled buffer in front of the constraint, so it is not left waiting for parts.

Check incoming quality there too, so it does not spend time on parts that are already bad.

4. Shrink batch sizes

Large batches create long queues, like the 480-minute wait before paint in the example.

Smaller batches shorten lead time and expose problems sooner. One-piece flow is the far end of this idea.

5. Make changeovers short

Plants often run big batches because changeovers are long. Cut the changeover first, then cut the batch.

The SMED method moves as much setup work as possible outside the stop.

6. Improve first pass yield at the weakest step

Start with the step that has the lowest yield, since it pulls RTY down the most.

Use statistical process control to spot drift early, and poka-yoke to make the common errors impossible.

7. Pull work instead of pushing it

Release work only when the next step can take it. That keeps WIP and queues under control.

A kanban system is a common, low-tech way to start.

8. Shorten travel and handling

Moving parts adds lead time and handling damage, but no value.

Put steps that follow each other close together, and remove double handling at storage points.

9. Standardize the work

Different shifts often run the same step in different ways, with different results.

Write down the best known method as standard work and train every shift to it.

10. Keep the bottleneck reliable

A breakdown at the constraint stops the whole flow. Put the bottleneck's known failure points on a preventive maintenance schedule.

Keep its critical spare parts in stock, so a repair does not wait for delivery.

Common mistakes

• Speeding up a step that is not the bottleneck. Output stays the same and the queue grows.
• Measuring cycle time and calling it lead time. The waiting between steps is often most of the lead time.
• Averaging step yields. Multiply them instead, or the process looks better than it is.
• Counting reworked units as first-pass good. Rework uses capacity at the steps it goes through again.
• Trusting capacity on paper. Plan with the OEE the bottleneck actually achieves, as the example showed.
• Measuring once. Process efficiency drifts, so review it on a regular rhythm.

A 30-day plan

1. Week 1: map one product family from material release to shipping, with every queue.
2. Week 2: time the steps, record first pass yield per step and confirm the bottleneck.
3. Week 3: start automatic data capture at the bottleneck and review its top losses every day.
4. Week 4: pick one queue to shrink and one loss to fix at the bottleneck, then measure lead time, RTY and bottleneck OEE again.

A short daily OEE meeting keeps the actions moving after the first month.

How Fabrico helps

Fabrico is an OEE platform with a full CMMS built in, which makes it a practical tool for the bottleneck part of process efficiency.

It collects machine data through PLC connections, IoT sensors and AI cameras, and shows availability, performance, quality and OEE in real time.

It detects micro stops and shows downtime, MTTR and MTBF by machine, so you can see where and when the constraint loses time.

Because maintenance lives in the same platform, your team can open work orders, run preventive maintenance plans and track spare parts for the bottleneck without switching tools.

With multi-plant views, you can benchmark OEE and downtime for the same machine type or line across sites.

Want to see what your bottleneck really delivers? Book a demo.