Manufacturing KPIs: Your OEE Went Up and Customers Got Angrier

The monthly review looks like a success story. The extrusion team ran longer batches, cut changeovers by a third, and OEE climbed four points. In the same month, three customer orders shipped late, and sales spent the review asking why the factory "got better" while customers got angrier.

Nobody cheated. The team optimized the number it was given. Manufacturing KPIs (key performance indicators) are the handful of measures a plant uses to judge whether it is doing well: equipment effectiveness, delivery, inventory, throughput, schedule adherence. Each one is useful. The trouble starts when one of them is read alone.

Why good KPIs fight each other

Think of a basketball team where every player is scored only on their own points. Everyone shoots, nobody passes, and the team loses with record individual stats. A factory works the same way. Each department controls one number, and many of those numbers pull in opposite directions.

Here are the four pairs that collide most often.

1. OEE up, OTIF down. Long batches mean fewer changeovers, so equipment effectiveness rises. But the urgent small order waits behind the long batch and ships late. How OEE is calculated, and why the same score can hide opposite problems, has its own article.

2. Throughput up, WIP up. Release more jobs to the floor and the first machines get busier. If the bottleneck can't absorb them, the extra work just waits between operations. Little's law makes this precise: lead time = WIP ÷ throughput. With 200 jobs in progress and 50 finished per day, the average job takes 4 days. Push WIP to 300 without raising the bottleneck's output, and it takes 6.

3. Inventory down, shortages up. Cutting stock looks great on the balance sheet until the one component you trimmed stops three orders. Inventory has to be judged by where it sits and what it covers, which is the subject of inventory management.

4. Schedule adherence up, deliveries still late. The floor followed the plan perfectly, but the plan put stock replenishment ahead of overdue customer orders, or was rebuilt every morning so adherence measured a moving target.

A toy example: three orders, one machine

One extruder, currently set up for blue. Each color change takes 2 hours. Three orders:

• Order A: blue, 8 hours, due at hour 16
• Order B: red, 4 hours, due at hour 6
• Order C: blue, 6 hours, due at hour 24

Plan 1, the OEE-friendly one. Stay on blue, run A and C, then change to red for B. A runs 0–8, C runs 8–14, changeover 14–16, B runs 16–20. One changeover. B was due at hour 6 and finishes at hour 20: 14 hours late. On-time delivery: 2 of 3.

Plan 2, the delivery-friendly one. Change to red first, run B, change back. Changeover 0–2, B runs 2–6, changeover 6–8, A runs 8–16, C runs 16–22. Two changeovers. B finishes at 6, A at 16, C at 22. All three on time.

Now the equipment view. Both plans have 18 hours of production. Plan 1 needs 20 hours on the machine (18 + 2), so availability is 18 ÷ 20 = 90%. Plan 2 needs 22 hours (18 + 4), so availability is 18 ÷ 22 ≈ 82%.

Plan 1 wins on OEE. Plan 2 wins with the customer. If the extruder is not your bottleneck, the extra 2 hours of changeover cost nothing real, and Plan 2 is simply the better plan. If it is the bottleneck, the trade-off is real and deserves a decision, not a default. Finding out which machine really limits output is the starting point of manufacturing optimization.

The core manufacturing KPIs, in one line each

• OEE: how much of the planned time a machine produced good parts at full speed. Availability × performance × quality.
• OTIF (on time, in full): the share of orders delivered on the promised date and in the full quantity. Watch out: 95% on time and 95% in full can multiply to far less; see OTIF.
• Inventory turns: cost of goods sold ÷ average inventory. $12 million of annual COGS with $2 million of average stock gives 6 turns, or about 61 days of cover (365 ÷ 6).
• Schedule adherence: jobs (or hours) completed as scheduled in a period ÷ jobs (or hours) scheduled.
• Throughput: good units completed per period at the end of the process, not at the first machine.

Myth: "More KPIs give you more control"

A dashboard with 40 metrics gives every department a green number to point at. Plants that manage well read a few measures in pairs that check each other:

• OEE together with OTIF
• Throughput together with WIP or lead time
• Inventory turns together with shortage count
• Schedule adherence together with the number of schedule changes

If one number in a pair improves while its partner gets worse, you have not improved. You have moved the problem.

A one-page KPI definition sheet

Most KPI arguments are really definition arguments. "Our OTIF is 92%." "Against which date?" Write one page per KPI and agree on it once:

The "date used" line alone settles half the arguments. If every missed promise quietly resets the due date, OTIF will look perfect forever.

How to review KPIs without blaming the wrong team

1. Start from one exception, such as a late order, not from an average.
2. Walk it back: was it released late, short of material, pushed behind a long batch, stuck at the bottleneck, or held by quality?
3. Check the pair: if OEE went up the same week, look at the sequence.
4. Fix the cause, then watch both numbers in the pair.

FAQ

What are the most important manufacturing KPIs? For most plants: OTIF for the customer, OEE at the bottleneck for capacity, inventory turns for cash, and WIP or lead time for flow. Always read them together.

What is a good OEE? An 85% figure is often cited as "world-class," but it is a benchmark, not a law. A machine that isn't the bottleneck doesn't need it.

How it looks in factory.online

In factory.online, the toy example above is two planning scenarios: the same three orders sequenced two ways on the finite-capacity Gantt, with due dates and changeovers visible side by side. Output and scrap confirmations on job orders supply the actual quantities behind OEE and throughput, and the Power BI integration puts those confirmations next to delivery data, so the pair is on one screen instead of in two departments' reports.

BEEDIGIT collects machine and operator data for OEE, downtime and actual production reporting. These records help the team compare planned capacity with what the equipment actually achieved and investigate recurring losses.

If your last KPI review ended with every team green and the customer red, book a demo and bring that month's numbers.