Value Stream Mapping

Value stream mapping traces the complete journey of a product family: every operation, every stock, every transport, and above all the information flows that trigger all of it. The method was popularised by Mike Rother and John Shook in Learning to See, published by the Lean Enterprise Institute.

Its strength is not the drawing, it is the timeline at the bottom of the map. On it are plotted, alternately, value-added time and waiting time. The ratio between the two is almost always brutal: a few minutes of value added for several days of lead time. It is that figure that triggers decisions, not the map itself.

A VSM is done in two stages: the current state, recorded on the shop floor with a stopwatch and a pencil, then the future state, which is a design decision. Without a future state, a VSM is only an expensive observation.

Cycle time, changeover time, machine availability, batch size, number of operators, scrap rate, and the stock present upstream — expressed in days of coverage rather than in parts, so as to be comparable from one workstation to the next.

The recording is done on the shop floor, never from the ERP. The theoretical times of the routing describe the process as it was designed; the VSM must describe the one that runs today, with its disruptions.

The VSM serves to settle a question that poisons workshops: should the operations be sped up, or should the waiting between them stop? As long as nobody has added up the waits, the default answer is to speed up the operations — because they are the only things one can see working. The map shows that the lead time is almost never decided there.

It also serves, and perhaps above all, to get people who do not talk to each other to look at the same object. Production, planning, logistics and purchasing each have a partial, consistent view of the flow, and those views do not contradict each other: they do not overlap. Half a day in front of a map drawn in pencil produces more agreements than a quarter of meetings, because the disagreement finally bears on a visible object.

One has to be clear about what it does not do. A VSM improves nothing: it measures and it points the way. What follows — pull flow, changeover reduction, rebalancing, grouping of operations — is not VSM, it is the projects the map has designated. An organisation that maps a lot and launches nothing has built a very pretty dashboard of its own inertia.

Seven operations, a downstream consumption of 480 parts per day. The value-added time recorded with a stopwatch over the whole journey totals 18 minutes. The stocks observed between the operations, counted the same morning, are 3,200, 1,800, 4,500, 900, 2,700 and 6,000 parts.

Converted into days of coverage — each stock divided by the 480 parts consumed daily — they give 6.7 · 3.8 · 9.4 · 1.9 · 5.6 · 12.5 days. Total: 39.9 days of lead time for 18 minutes of real work. The ratio is in the order of one thousandth: over an eight-hour day, the part is being transformed for less than a minute.

It is the conversion into days that does the work. Saying "6,000 parts in front of dispatch" provokes no reaction in a workshop used to those volumes; saying "twelve and a half days" provokes an immediate one, because the figure is at last comparable to the lead time promised to the customer.

Then comes the reading that changes the action plan. The two biggest stocks are not in front of the slowest machines. The 6,000 parts are waiting because the carrier collects weekly; the 4,500 are waiting because final inspection runs on one shift while the line runs two. Twenty-two days out of forty come down to two organisational decisions, neither of which costs an investment to correct.

That result is commonplace, and that is what makes it useful: the map almost never reveals a technical problem. It reveals rules — collection frequencies, batch sizes, staggered hours, trigger thresholds — that were decided once, often for good reasons since lapsed, and that nobody discusses any more because nobody sees them any more.