A value stream map your scheduler will actually use.

Walk into most shops that have done a value stream mapping event and you will find the map, laminated on a wall near the break room. Now ask the scheduler what they used to build this week's plan. It will be a spreadsheet, the ERP dispatch list, and a stack of hot-list emails. The map and the schedule have nothing to do with each other.

A value stream map is supposed to be the blueprint for how work flows. The schedule is how work actually flows. If the two never meet, the map is decoration. Here is how to draw one the scheduler will run, and what to do with it on Monday.

What a value stream map is, and what it is not

The Lean Enterprise Institute defines value stream mapping as "diagraming every step involved in the material and information flows needed to bring a product from order to delivery."^[1] The tool came from Toyota, which calls it a material and information flow diagram.^[1] Mike Rother and John Shook's workbook Learning to See, published by LEI, introduced the tool to thousands of people and received a Shingo Research Prize in 1999.^[2]

Two parts of that definition matter. First, "information flow." LEI's description of the workbook says the map "differs significantly from tools such as process mapping or layout diagrams because it includes information flow as well as material flow."^[2] A process map shows where parts go. A value stream map also shows who tells each process what to make next, and when. That second half is the scheduler's whole job.

Second, "order to delivery." The map is door to door for one product family, not one department. LEI's starting list: select a product family, have one person lead, begin door to door, and cover both material and information flow.^[1]

NIST MEP lays the method out as four steps: pick the product value stream, draw the current state of materials, processes, and information flow, diagnose problems and define a future state, then implement.^[3] Most shops do the first two well, the third badly, and the fourth not at all.

One clarification from John Shook matters for schedulers: "VSM does not equal 'flow analysis' or the process of designing and creating optimum product flows."^[4] It is a tool for seeing how flows operate. The scheduling logic comes after, and it comes from the future state.

Current state: walk it, do not export it

The current state map is built by walking the floor with a pencil, not by exporting the routing file. Shook's reason for hand drawing is that it forces you to go look at the whole stream, not one process.^[4]

I spent 13 years fixing late orders by walking floors and asking why. The pattern was the same everywhere: good people, scattered data, nobody looking at the same numbers. The current state map is often the first time they look at the same picture.

Pick one product family

LEI's lexicon defines a product family as "a product and its variants passing through similar processing steps and common equipment just prior to shipment to the customer," and notes that product families are the unit of analysis for value stream maps.^[5] Aaron Hunt, writing on lean.org about low-volume, high-mix plants, describes the job shop fix: group products by how they flow, not what they are. His teams defined families like "Lathe-Vertical Mill-Vertical Mill," and the parts in each "might have been apples vs. oranges vs. pears in appearance, but if they all used the same flow, that was our starting point."^[6]

A routing family is a schedulable thing. "All the stainless parts" is not.

Collect the data boxes the scheduler needs

Under each process box goes a data box. LEI's list is the same list a scheduler needs for finite capacity: cycle time, changeover time, uptime, percent complete and accurate, and available time per shift.^[1] Measure them. Do not copy the standard from the quote.

Michael Ballé warns that trying to map every component in a product "is like digging a hole in the sand: the more you dig, the more you find."^[7] Map the main flow of the family. Note feeder parts, but do not chase them on day one.

Count the piles and draw the information flow

Between process boxes you record the inventory waiting. Rother's point is that piles tell you where flow is interrupted by a process problem, and the question at each one is why the shop holds that much there.^[8] Each pile is a queue the scheduler already manages by expediting. Write down the real reason. It becomes a scheduling rule later.

Ballé's view is that information flow is where the map is most useful, and most teams sketch it as a dotted line from logistics to the cell. His questions: how does information arrive from the customer, how often is it sent to each process, and where do customer instructions become production instructions.^[7] This is where a small shop discovers three people are sending three priority lists to the same work center.

Future state: answer the questions, not the waste list

Drew Locher, writing for LEI, says "50 percent of the maps I am asked to review are just current states with no true future states." They are a current state covered in kaizen bursts, moved to a list, prioritized, assigned, and closed out as an event. He calls it a war on waste.^[9]

The alternative is to answer the future-state questions in Learning to See. The first is what the customer really needs, which sets the demand rate, lead time target, and quality target.^[9] LEI's lexicon lists the guidelines the rest walk through, and each is a scheduling decision.^[1]

Takt in a job shop

Takt time is available production time divided by customer demand.^[10] The job shop objection is that demand is lumpy and every part is different. Shook's answer is that the idea takt has no place in low-volume job shops is a misconception, that most of Toyota's early suppliers were job shops, and that takt is most useful where flow is hard to see.^[4] For a family, takt is the rate it must leave the building to keep up with orders. Toyota reviews takt monthly, with a tweaking review every ten days.^[4]

Pacemaker, not bottleneck

LEI defines the pacemaker process as "any process along a value stream that sets the pace for the entire stream," usually near the customer end, and adds that it should not be confused with a bottleneck.^[11] Rother and Harris, in Creating Continuous Flow, say every facility has at least one pacemaker, usually where products take their final form, and that how you run it determines both customer service and the demand pattern on every upstream process.^[12]

Why not schedule the bottleneck? Shook's answer: lean systems work to break bottlenecks, but a bottleneck exists for problematic reasons such as downtime, poor quality, or long changeovers, and he will not let an operation with those problems dictate the pace of the whole flow.^[4] You fix the bottleneck. You schedule the pacemaker. If you are not sure which your busiest machine is, our post on why 85% machine utilization can still mean late orders covers the difference.

Once the pacemaker is chosen, LEI's guideline is that with pull in place you typically schedule only one point in the value stream.^[1] The scheduler stops planning twelve work centers and plans one.

Why most maps die on the wall

Jim Womack wrote an LEI column resolving that nobody would turn their maps into corporate wallpaper by failing to implement the future state.^[13] His diagnosis of why they do:

"But much more often the problem is simply that there is no effective management of each value stream."^[13]

The maps were done by an improvement group as a special project, and nobody took responsibility as a day-to-day manager for implementing the future state and treating it as the new current state.^[13]

Three other failure modes repeat in the lean literature.

The future state is too far away. Rother's rule: more than about six kaizen bursts on a future state map means you are getting ahead of yourself. Sketch the two-to-three-year direction, then set a target condition no more than 12 months out.^[8]

The wrong people, and no plan. Judy Worth, an LEI faculty member, names the typical mistakes as not having the right people involved and not generating a plan to get from current to future state.^[14] If the scheduler was not in the room, do not expect the scheduler to run the result.

The management never changed. Locher observes that shops change the workflow but not how the stream is managed, which makes reversion likely.^[9]

The authors saw this coming. In the foreword to the sequel, Womack, Jones, and Shook warn that "anyone can draw an attractive future state on paper."^[15]

Turning the future state into what the scheduler runs

The fix for all four is the same: the future state has to become the scheduler's working structure, not a poster. It becomes three things.

1. Routings

Each process box on the future state map becomes an operation on the family's routing, carrying the data box numbers: measured cycle time, changeover, uptime, and first-pass yield.^[1] Where the future state merged operations into continuous flow, the routing shows one operation with no queue allowed between. Where the current state had a hidden rework loop, the routing shows it.

2. Stages

The future state map breaks the stream into segments: a continuous-flow cell here, a FIFO lane there, a supermarket between. Each segment is a stage, and each needs one rule the scheduler can enforce. LEI describes the FIFO lane rule exactly: the supplier fills from the upstream end, the customer withdraws from the downstream end, and "if the lane fills up, the supplying process must stop producing until the customer consumes some of the inventory."^[16] That is a WIP cap. A supermarket is a min/max. A continuous-flow cell is a stage with zero allowed queue.

Stages are also where visibility lives. If the system tracks which stage every job is in, the scheduler sees the value stream on a screen. HixonOps builds routings and value stream stages into its MES for this reason: the map becomes the structure the floor reports against, tied to the ERP you already run.

3. Pacemaker scheduling

The scheduler releases work at exactly one point: the pacemaker. LEI's guidance is to release and withdraw small, consistent increments there, usually 5 to 60 minutes' worth, a practice it calls paced withdrawal; the increment is the pitch.^[1] In a job shop the pitch is more often a few hours or a shift, but the principle holds: steady, small releases at one point instead of a weekly dump of work orders onto every work center.

Downstream of the pacemaker, jobs run FIFO. Upstream, operations replenish what the pacemaker consumed. True one-off components (what Ballé calls strangers) are still scheduled the old-fashioned way, with production instructions sent straight to the queue and zero safety stock.^[7]

What it looks like when it works

MEP center case studies are short and the results are reported by the client, so treat them as examples, not benchmarks. But they show the same pattern: map the current state, design a future state around flow, then change how work is released.

Tank Technology, a Wisconsin tank maker with about 50 employees, wanted to go from six or seven units a day to a consistent eight. WMEP mapped the product family and moved the work into a cell; the company reported lead time cut from 14 days to 4 and backlog down 60%.^[17]

QCast Aluminum, a Pennsylvania foundry, mapped why production could not meet delivery dates and found finishing WIP in scattered locations. The future state set explicit targets (7-day throughput, 4-week lead time, 99% on time), and the first target condition was a supermarket holding finishing WIP at 3 to 5 days.^[18] That is a stage with a rule.

At REELEX Packaging in New York, the event ended the way every one should: "Upon completion of the future state map an action plan was created with specific improvement activities for the team to implement."^[19]

In my own work, one turnaround I come back to did not come from a new system. A steel products maker had over $700,000 tied up in late orders. Six months later it was under $50,000. What changed was that everyone looked at the same numbers every day. That is what a map is for.

Keep it alive

WMEP's infosheet says it plainly: value stream mapping "is not a one-time event."^[20] Rother's version is that once a target condition is reached you need the next one, or performance slips back.^[8]

If the scheduler runs the pacemaker every day, the stages show on a screen, and the routings carry measured numbers, the map does not need to be on the wall. It is in the schedule. Rother and Shook wrote that "wherever there is a product for a customer, there is a value stream. The challenge lies in seeing it."^[21] The scheduler sees it every morning. Build the map so they can use it.