Measuring the Right Thing

The first measure a facility adopts is usually units per hour per person. It is easy to compute, easy to compare, and it explains very little.

Why output per person misleads

It attributes system performance to individuals. A picker in a badly slotted aisle is slower than one in a good aisle, and the difference is the layout.

It ignores what the work was. Fifty small items and fifty pallets are not comparable.

It ignores waiting, which is frequently the largest component of elapsed time and is not the worker's doing.

It encourages the wrong behaviour: skipping checks, unsafe handling, cherry-picking easy work.

It creates a number that looks actionable when the action available is almost always to change the process.

The measures that explain

Waiting time, by stage. Where does work sit between being ready and being started.

Travel distance and time per task, which is the largest controllable loss in most picking operations.

Touches per unit. How many times an item is handled between receipt and despatch. Every touch is cost and risk.

Variation, not just averages. A stage with a stable mean and huge spread is a different problem from a slow stable one.

Exception rate, and where exceptions originate rather than where they surface.

Equipment availability against demand for it.

Each of these points at a change the operation can make, which is the test of a useful measure.

The elapsed time decomposition

The single most informative analysis available, and it requires only timestamps you probably already have.

Take a unit of work from arrival to completion.

Split the elapsed time into: waiting, travel, handling, checking, and rework.

Express each as a proportion.

Handling is usually a minority of elapsed time, frequently a small one. Operations that discover this stop trying to speed up handling and start removing waiting.

Aggregation, not attribution

Almost every question worth asking is answerable without naming anyone.

Where is the bottleneck. Why is the afternoon slower. Did the new layout help. Which product lines cause the most rework. Where do errors originate.

All aggregate questions. None requires individual measurement, and answering them aggregately avoids the entire category of problems covered in the worker monitoring section.

Where the data usually already is

Scan timestamps at each handling point.

Task issue and completion events from the management system.

Equipment telematics.

Order data, which brackets the whole flow.

Join these before installing anything. Most facilities can produce the elapsed time decomposition from existing records within a week, and that analysis frequently changes the priorities enough to redirect the whole project.

The first question to answer

Where does work wait?

If the answer is unknown, that is the project. If it is known and nothing has been done about it, monitoring is not the constraint.

The touches count

A simple measure that reliably finds waste and is rarely computed.

Count how many times an item is handled between arrival and despatch.

Every touch is time, cost and damage risk.

Compare across product lines. The variation is usually large and unexplained.

Look for double handling: staging that could be eliminated, a putaway that could go direct, a consolidation step that exists because of a system limitation nobody has revisited.

Reducing touches is one of the few changes that improves cost, speed and quality simultaneously, which makes it worth finding before anything else.

Choosing what to report weekly

Five measures, chosen so that each points at a different decision.

Waiting time by stage, as a proportion of elapsed time.

Cycle time median and ninetieth percentile.

Travel per order line.

Exception rate by origin.

Equipment availability against demand at the peak hour.

With the normal range marked on each, so ordinary movement does not generate questions.

One page. A dashboard with twelve tiles produces no attention on any of them.