How to do value stream mapping, step by step

To do value stream mapping, walk the process, record data at each step and draw the current state by hand.

Value stream mapping is a widely used lean tool, and an easy one to misuse. Teams often draw it in a meeting room as they believe the process works. The result looks tidy but can be far from reality.

Mike Rother and John Shook's book Learning to See describes walking the process and drawing the first map by hand. This guide follows that approach.

Value stream mapping vs process mapping

Is this stream worth mapping? Six checks

Who should be in the room

Step 1: walk the process

Choose one product family and walk its path from the customer back to the supplier on a gemba walk. A clipboard, a pencil and a stopwatch are all you need.

Walking backwards, from shipping to receiving, keeps attention on what the customer pulls rather than what each department pushes.

Step 2: collect data at each step

Record cycle time, changeover time, uptime and the number of people at each step. Count the inventory waiting between steps, and convert it into days of demand.

Note how each step learns what to make next, such as a schedule, a forecast or a kanban signal.

Each process box on the map gets a data box. These are the fields most teams fill in, and how to get each one on the walk:

Cycle time (C/T): the time between one finished part and the next at that step. Time 10 parts in a row with a stopwatch and take the typical value, not the fastest.

Changeover time (C/O): the time from the last good part of one product to the first good part of the next. Ask when the next changeover is and watch it, or ask operators and check against a log.

Uptime: the share of planned time the step is actually able to run. Use the machine log if there is one, otherwise ask about the last week's stops and estimate.

Operators: the number of people the step needs per shift.

Batch size or EPE: how many parts run between changeovers, or how often each part number is made (every part every day, every week and so on).

Inventory: count the parts waiting before the step, then divide by daily customer demand to turn it into days. For example, 1,000 parts waiting and demand of 400 a day is 2.5 days.

A simple walk sheet keeps the data in order. Make one row per step with blank columns for step name, C/T, C/O, uptime, operators, batch size, inventory count and inventory days, plus a notes column for how the step gets its instructions. Fill it on a clipboard as you go, then draw the map from it.

Inventory is often the largest number on the map, and it is a large number across the economy too. US businesses held $2,818.8 billion of inventory at the end of August 2026, about 1.33 months of sales, according to the Census Bureau.

Step 3: draw the current state

Sketch the map on a large sheet with the people who do the work. Add the information flow, such as orders and schedules, above the material flow.

Then add the timeline at the bottom: value-added time against total lead time. The difference shows how much of the lead time is waiting.

A current-state map: an illustrative example

Here is a small made-up example of the numbers a current-state map holds. A plant makes one family of steel brackets. Customer demand is 400 brackets a day, and the plant runs one shift of 450 working minutes.

Raw steel waits before the first step: 6.0 days.

Cutting: C/T 15 s, C/O 30 min, uptime 95%, 1 operator. Then 2.5 days of cut blanks wait.

Bending: C/T 30 s, C/O 45 min, uptime 90%, 1 operator. Then 3.0 days wait.

Welding: C/T 120 s, C/O 10 min, uptime 85%, 2 operators. Then 2.0 days wait.

Painting: C/T 45 s, C/O 60 min for a color change, uptime 92%, 1 operator. Then 1.5 days wait.

Assembly and pack: C/T 102 s, no changeover, 2 operators. Then 3.5 days of finished goods wait to ship.

The timeline at the bottom of the map adds the two rows. Value-added time is 15 + 30 + 120 + 45 + 102 = 312 seconds. Lead time is 6.0 + 2.5 + 3.0 + 2.0 + 1.5 + 3.5 = 18.5 days. A bracket spends about 5 minutes being worked on and more than 18 days waiting.

Marking problems on the map

Step 4: design the future state

A current-state map is only useful if it leads to a future state and a plan. Learning to See offers a set of questions for this.

Ask where you can create continuous flow and where you need a pull system with kanban. Ask which single point to schedule, and how to level the mix with heijunka.

Turn the ideas into improvement loops with owners. Long changeovers on the map often point to SMED, and takt from the takt time calculator sets the pace.

Learning to See frames the future state as a short list of questions. Here they are in our own words, answered for the example map above.

What is the takt time? 27,000 seconds a day divided by 400 brackets is 67.5 seconds. Welding at 120 seconds and assembly at 102 seconds are both slower than takt for a single station, so each needs two stations working in parallel or less work content.

Will you build to a finished goods supermarket or ship directly? With steady demand, a small supermarket of about 1 day can replace the 3.5 days now held.

Where can you use continuous flow? Welding and assembly sit next to each other and have similar cycle times, so they can become one cell with no stock between them.

Where do you need supermarket pull? Painting batches by color, so a small supermarket after painting, refilled by kanban, keeps it from pushing stock onto the cell.

Which single point will you schedule? The welding and assembly cell. Every step before it is refilled by pull, so only one schedule goes to the floor.

How will you level the mix? Paint and build a little of each color every day instead of one color per week. That needs shorter color changes, which points to SMED on painting.

What process improvements do you need? SMED on painting and bending, and work on welding uptime. Each one gets an owner and a date on the plan.

Turning the map into an action plan

Common value stream mapping mistakes

Mapping from the meeting room. A map drawn from memory shows the process as people believe it runs. Walk it, and take the numbers you see, even when they are awkward.

Mapping too many product families. One map should follow one family that shares most steps. Mapping everything at once gives a map so busy nobody can read the waste.

Averaging the inventory away. Count what is actually there on the day you walk, and note if the day looks unusual, rather than using the target levels from the planning system.

Leaving out the information flow. Half the delay is often in how orders and schedules reach each step. Draw it above the material flow.

A future state with no owners. A future-state map with no plan, names and dates is a nice drawing. Each change needs an owner and a review date.

Mapping office and service work

The same method works for requests, approvals and handoffs. Inventory becomes items waiting in queues or inboxes, and the timeline still compares working time with waiting time.

Waiting for information is part of that queue time. The McKinsey Global Institute estimated that interaction workers spend nearly 20% of their time looking for internal information or tracking down colleagues who can help with a task.

Office maps usually add three measures to each data box:

• Percent complete and accurate: how often work arrives at a step ready to use, without chasing missing information.
• First-pass yield: how often a step finishes without rework.
• Handoffs: how many times the work changes hands between people, teams or systems.

An illustrative service example: a request arrives through a web form, waits in a shared inbox, goes to a specialist, goes back to the customer for missing details, waits again, then gets reviewed and approved. Each active step is short. The map shows why the total takes days: many requests arrive incomplete, approvals are batched twice a week, and specialists cannot see which requests are urgent. The future state adds required intake fields, a visible priority queue and daily approvals.

Where software helps

Software earns its place after the walk. It keeps maps current, calculates lead times, shares maps across sites and compares versions over time.

Some tools add simulation to test a future state, or process mining to rebuild the flow from system data. See value stream mapping software. Software cannot do the observing.

Frequently asked questions