Lean tools

Work in progress (WIP)

In French: En-cours (WIP)

Work in progress is everything that has entered a process without yet having left it: started parts, open files, unclosed tickets. It is the quantity that is easiest to count and most rarely steered.

What the term covers

Work in progress is neither a stock of raw materials nor a stock of finished goods: it is what lies between the two, committed and unfinished. In a workshop, it is the bins waiting in front of the stations. In a department, it is the open files, the requests taken on and the tasks started.

Its particularity is that it is almost always invisible in the dashboards of administrative activities. Nobody counts open files, whereas everyone would count pallets; yet the quantity is the same and produces the same effects.

A high level of work in progress always gives an impression of intense activity, and that is what makes it hard to fight. A lot of work started looks like a lot of work done, whereas it is exactly the opposite: nothing is finished.

What it costs

It lengthens the lead time. It is its most direct and least known consequence: at unchanged output rate, doubling what waits in the process doubles the time each request takes to cross it.

It delays the discovery of defects. A problem that appeared at one step is detected only at the next; the more work waiting between the two, the higher the number of copies already produced with the same defect at the moment it is found.

It ties up money. Each committed unit has consumed material, time and sometimes subcontracting, without yet having invoiced anything.

It hides problems. A station fed by a comfortable reserve never stops, even when its supplier fails. The reserve absorbs the failure and makes it invisible, hence lasting — which is why one speaks of a water level that is lowered to reveal the rocks.

Limiting it, and at what level

Limiting consists of setting a cap on the number of units committed at the same time, and launching nothing new as long as the cap is reached. The mechanism is the same in a workshop with cards, on a visual board with a number per column, or in a department with an intake rule.

The effect is immediate and counter-intuitive: the queue does not disappear, it moves upstream, where it is visible and costs nothing. What has entered the process moves fast; what has not entered waits its turn in full view.

The right level cannot be calculated in advance. You start from what exists, lower the cap in small steps, and watch what breaks: it is precisely what breaks that points to the next problem to treat. Going down abruptly reveals everything at once and blocks the process, which makes people abandon the approach.

One point deserves to be stated from the outset with the teams: this cap reduces the work committed, never the work delivered. The confusion between the two is the main source of resistance, because a less loaded board gives the feeling of doing less.

The common mistakes

Launching more to go faster. It is the natural reflex when behind schedule, and it produces the opposite: every additional launch lengthens the queue that every request will have to cross, including those one was trying to speed up.

Setting the cap at the current level so as not to upset anyone. A cap that never bites changes nothing; its point begins exactly where it forces finishing before starting.

Confusing high work in progress with high workload. Many organisations saturated with work in progress are not short of means: they simply open faster than they close, and nobody looks at the gap between the two.

On the shop floor — Software development

A team of six was tracking twenty-nine subjects open at the same time. The rule of not exceeding two subjects per person produced, in the first week, a visible queue and a lot of tension. By the fourth, the number of subjects closed per week had risen without any other working method having changed.

The other tools illustrated by this trade: services, it and engineering.

Frequently asked questions

How do you count the work in progress of an administrative process?

By counting the requests that have an entry date and not yet an exit date, at a given moment. Most management tools allow that count without any special development, and it is enough to record it every week on a fixed day and time. The difficulty is not technical: it is that nobody thinks of looking at that quantity, never having seen it in a dashboard.

Does limiting work in progress not amount to deliberately slowing down?

No, and it is the main objection to address. The limit bears on what is started, not on what is finished: the output rate remains at least identical, and generally rises because interruptions and context switches decrease. What visibly slows down is the intake — and that is precisely what is sought.

Should there be an overall cap or a cap per step?

A cap per step is more effective because it points to the exact place where work accumulates, whereas an overall cap lets the imbalance form inside. Starting with an overall cap remains useful when the steps are not yet clearly separated: it already produces most of the effect on lead time, and the split comes afterwards.

Is zero work in progress the goal?

No, a process needs a minimum of committed work so as not to stop at every disruption, and aiming for zero would make the slightest irregularity blocking. The goal is the lowest level at which the process still holds, which assumes having first reduced the disruptions that justified the previous level.