Line Balancing: How to Balance Work to Takt Time (With Yamazumi Chart)

By Linh Ngo · Updated Oct 10, 2026

Line balancing divides a process's work into stations so each station's total task time fits under takt time, with as little idle time as possible. It lets a line meet customer demand with the fewest people or stations, and it shows the real bottleneck on a single chart instead of in an argument.

What Is Line Balancing?

Line balancing is the act of assigning tasks to workstations (or people, or queues) so that the workload is as even as the sequence of the work allows. Every station gets about the same amount of work, and no station gets more than the pace of demand permits.

That pace is takt time: net available time divided by customer demand. If takt is 100 seconds, the goal is to arrange the work so that no station needs more than 100 seconds per unit. And since idle time costs money just like overload costs quality, the second goal is to keep every station close to that 100.

Researchers call the textbook version the simple assembly line balancing problem: given tasks with fixed times and a required order, assign them to a sequence of stations. In one formulation, you minimize the number of stations for a given cycle time. In the other, you minimize the cycle time for a given number of stations. Ritt and Costa describe both versions and note that the decision versions are NP-complete, which is why real lines are balanced with rules of thumb and software rather than a tidy formula.

You don't need the math to get most of the value. You need three numbers (takt time, total work content, and the time at each station) and a willingness to move work around.

Where it applies

Assembly lines are the origin, but the logic holds anywhere work moves through stages in sequence: a claims team, a hiring pipeline, a support queue, a code review flow. If a stage runs hot while another waits, you have a balancing problem.

Line Balancing vs Related Concepts

People mix these up, so here's how they differ and fit together.

The short version: takt time sets the target, line balancing arranges work to hit it, and standard work keeps it there.

The Core Formulas

Four calculations cover almost every balancing exercise. The formulas below follow the standard textbook versions laid out in an ASEE case study on assembly line balancing.

Cycle time (target) = total available production time / number of units required. In lean terms this is takt time.

Theoretical minimum number of stations = total work content / cycle time, rounded up to a whole number. Total work content is the sum of every task's time.

Balance delay = (stations x cycle time - total work content) / (stations x cycle time), as a percentage. It's the share of paid capacity that sits idle.

Line efficiency = 100% - balance delay. Equivalently, total work content divided by (stations x cycle time).

A quick illustration. Say total work content is 290 seconds and takt is 100 seconds. The theoretical minimum is 290 / 100 = 2.9, so you need at least 3 stations. With 3 stations, capacity is 300 seconds, balance delay is 10 / 300 = 3.3%, and efficiency is 96.7%. That's close to the best you can do. Add a fourth station and efficiency drops to 72.5% (290 / 400), even though no one worked less hard.

Notice what the minimum number tells you. It's a floor, not a promise. Precedence rules and indivisible tasks usually push the real answer higher, and the gap between the floor and your actual count is the improvement opportunity.

Precedence Constraints: The Part That Makes It Hard

If tasks could be assigned in any order, balancing would be arithmetic. They can't. You can't paint a part before it's welded, can't approve an invoice before it's coded, can't ship a build before it's tested.

A precedence diagram records these dependencies. Each task is a node, and arrows show what must finish first. When you assign tasks to stations, a task can only go to a station if all its predecessors are in that station or an earlier one.

This is why a perfectly flat line is rare. Three things get in the way:

• Order. A long task late in the sequence can't move to a station with spare time upstream.
• Indivisibility. A 45-second task won't split neatly into 30 and 15 across two stations without redesigning the work.
• Skill and equipment limits. Some tasks can only happen at a particular machine or by a certified person.

Good balancing often means changing the work itself: splitting a task, moving a fixture, adding a template so a step gets faster, or re-sequencing steps that only looked dependent out of habit.

Heuristics for Assigning Tasks

For small problems you can balance by eye. For larger ones, practitioners use simple assignment rules. Two of the best known are covered in the ASEE case study, which also covers a third, the Kilbridge and Wester method.

Largest candidate rule

List tasks from longest to shortest. To fill a station, scan down the list and take the first task that satisfies precedence and still fits in the remaining time. Repeat until nothing fits, then open the next station. It's easy to run by hand, though it can favor big tasks sitting late in the sequence.

Ranked positional weight

Helgeson and Birnie's 1961 technique, as the same ASEE paper describes it, gives each task a positional weight: its own time plus the times of all tasks that follow it in the precedence diagram. Rank tasks by weight from largest to smallest, then fill stations in that order, always respecting precedence and the cycle time. Tasks near the start of a long chain carry big weights, so they get placed early, which is the sensible instinct.

Neither rule guarantees the best answer. They produce good, not provably optimal, solutions. If the results look uneven, try both and keep the better layout.

How to Balance a Line, Step by Step

1. Calculate takt time. Net available time divided by demand, using the takt time formula. Subtract breaks and planned stops.
2. List every task with a time. Use observed times from the floor, not what the procedure says. A time and motion study gives you reliable numbers, and repeat observations beat a single stopwatch pass.
3. Draw the precedence diagram. Ask of each task: what must be done before this can start? Challenge any dependency that exists "because we've always done it that way."
4. Compute the theoretical minimum stations. Total work content divided by takt, rounded up.
5. Assign tasks. Use the largest candidate rule or ranked positional weight. Fill each station as close to takt as you can without crossing it.
6. Plot the result on a Yamazumi chart. Stacked bars per station against a takt line. Look for bars over the line and bars far below it.
7. Rebalance by hand where the chart looks wrong. Move small tasks across the boundary, shorten the biggest task, or split it.
8. Trial it, then write it into standard work. A balance that only exists on paper isn't a balance.
9. Recheck whenever takt changes. Demand moves, and with it the target. A line balanced for 100 seconds is unbalanced at 80.

Worked Example

Say a team assembles a small device. Net available time is 27,000 seconds per shift and demand is 270 units, so takt is 100 seconds. The tasks, times and predecessors are below.

Total work content is 290 seconds, so the minimum is 3 stations (290 / 100 = 2.9).

A first attempt (grouped by whoever's nearby):

Four stations, and the line runs at the pace of station 1: 120 seconds per unit, which misses demand. Efficiency using the real cycle time is 290 / (4 x 120) = 60.4%.

After ranked positional weight. Rank by weight: A, C, B, D, E, F, G, H, I, J. Fill station 1 with A (30), then C (20), then B (40) for 90 seconds. D (50) won't fit in the 10 seconds left, so station 2 takes D, E, F for exactly 100. Station 3 takes G, H, I, J for exactly 100.

Three stations, every load at or under takt, efficiency 290 / 300 = 96.7%. The largest candidate rule lands on the same assignment for this small network. On bigger networks, the two rules will often disagree, and that's your cue to compare.

One station fewer, a line that meets demand, and no one told to work faster. That's the pattern to look for. The people didn't change, the arrangement did.

The Yamazumi Chart

Yamazumi is Japanese for "pile" or "stack," as the Lean Enterprise Institute lexicon notes. The Institute lists it as another name for the operator balance chart, a visual aid for building continuous flow in processes with several steps and several operators.

Here's how it reads. Each operator or station is a vertical bar. The bar is made of stacked segments, one per work element, with each segment's height equal to that element's time. A horizontal line marks takt time. The LEI describes the aim as getting each operator's total close to, but a little under, takt, which keeps the number of operators as low as possible.

What you look for on the chart:

• A bar above the line. That station can't keep pace. Move something off it.
• A short bar. That station has spare time. It can absorb a task from a neighbor.
• A tall segment. One element is eating most of a station's time. Can it be shortened, split, or automated?
• Segment colors. Many teams color-code value-adding work, necessary non-value-adding work (walking, setup), and pure waste, so rebalancing and waste removal happen in the same pass.

You can draw it in a spreadsheet as a stacked column chart, or on a whiteboard with sticky notes cut to scale. Cutting task cards to scale and shuffling them physically works well in workshops, because the group argues about the real constraints (can this go before that?) instead of the numbers.

Yamazumi table template

Copy this into a spreadsheet, fill in task times, then insert a stacked column chart with stations on the horizontal axis and add a line series for takt.

Two things make it useful. First, the "Gap to takt" column turns the chart into a to-do list: positive gaps are problems, large negative gaps are donors. Second, the "Type" column tells you whether to balance the work or eliminate it.

Line Balancing Across Business Functions

The same logic moves off the factory floor. Here are illustrative cases (the numbers are made up to show the method).

Operations and manufacturing. A packaging line has a labeling step at 85 seconds and a boxing step at 40, with takt at 70. Moving the carton-erecting task from labeling's neighbor to boxing, and giving the label step a pre-staged roll, brings both under takt.

Customer service. A support team triages, resolves and documents tickets. If takt works out to 15 minutes per ticket but documentation takes 12 on top of a 9-minute resolve, the documentation step is the hot spot. Shifting templated documentation to a lighter-loaded role rebalances without adding headcount.

Sales. A prospecting flow runs research, outreach and follow-up across three people. If research takes triple the time of the others, leads queue before outreach and the outreach rep idles. Pre-built account lists or shared research templates move work off the slow stage.

HR. Onboarding often has stages owned by HR, IT and the hiring manager. One stage with a four-day queue sets the pace for the entire new-hire experience. Mapping task time per owner and moving the account-setup checklist earlier evens the load.

Finance. Invoice processing runs through capture, coding, approval and payment. If coding takes the longest and approval waits behind it, splitting coding by vendor type or adding auto-coding rules for repeat vendors rebalances the flow.

Software. A delivery flow of build, code review, QA and release has the same shape. If review is where work waits, pair rotation or smaller pull requests can pull the review step under takt. This connects to single-piece flow: small batches expose imbalance quickly.

In knowledge work, task times vary much more than on an assembly line. Balance to the average with some buffer, and treat takt as a planning target rather than a stopwatch.

Do's and Don'ts

Do:

• Use observed task times from the actual work, not the procedure on paper.
• Start from takt time, not from the current headcount.
• Plot the result on a Yamazumi chart before changing anything.
• Challenge precedence rules that nobody can justify.
• Rebalance when demand, product mix or staffing changes.
• Pair balancing with heijunka so mix and volume don't swing the loads every hour.

Don't:

• Treat the bars as individual speed targets. The chart measures the design of the work, not the people.
• Balance to 100% of takt. Leave a little room for normal variation, or any hiccup starves the next station.
• Balance a line whose stations are unreliable. If a machine goes down twice a shift, fix that first or the balance is fiction.
• Move work between stations without updating standard work.
• Chase a fractional station. If the theoretical minimum is 2.9, the answer is 3, not an argument about 0.1.
• Optimize one balanced section while a bigger constraint elsewhere limits the whole system. The theory of constraints says to find that constraint first.

Common Mistakes

Using the wrong takt. If takt is calculated from gross rather than net time, every station looks like it has slack that doesn't exist.

Confusing average with worst case. A station that averages 95 seconds but swings between 80 and 120 will block the line on its slow days. Look at the spread, not just the mean.

Forgetting walking and setup. Time spent fetching material or changing tools is real load. It belongs on the chart.

Balancing once and walking away. New products, new tools, and new demand levels break a balance. Revisit it on a schedule.

Ignoring quality. Squeezing stations tight can push people to skip checks. A balanced line with rising defects hasn't improved anything.

Key Facts

• Line balancing aims for each station's workload to be close to, but slightly under, takt time, according to the Lean Enterprise Institute.
• The simple assembly line balancing problem comes in two forms: minimize stations for a given cycle time, or minimize cycle time for a given number of stations; the decision versions are NP-complete (Ritt and Costa).
• Theoretical minimum stations = total work content / cycle time, rounded up; balance delay = (stations x cycle time - work content) / (stations x cycle time) (ASEE case study).
• Ranked positional weight, developed by Helgeson and Birnie in 1961, sums each task's time with the times of every task that follows it (ASEE case study).
• Yamazumi means "pile" or "stack" in Japanese (Lean Enterprise Institute).

Related Reading

• Takt time: the target every station's load has to fit under
• Standard work: how to lock in a balance once you've found it
• Time and motion study: getting reliable task times to balance with
• Theory of constraints: finding the one step that limits the whole system
• Value stream mapping: choosing which part of the flow to balance first