Line balancing calculation: stations, efficiency, and balance delay, worked through

Line balancing means assigning an assembly's tasks to stations so that no station's work exceeds takt time, using as few stations as possible. You need four numbers: takt time (the pace demand requires), the theoretical minimum number of stations, line efficiency, and balance delay (the idle time efficiency leaves behind). All four come from task times and one division each. The hard part is the assignment in between, because tasks have an order.

Here is every formula, applied to one line from start to finish.

The formulas

Takt time = available production time ÷ demand

Minimum stations = ⌈ sum of task times ÷ takt time ⌉ (round up)

Line efficiency = sum of task times ÷ (number of stations × cycle time)

Balance delay = 1 − line efficiency

The ⌈ ⌉ brackets mean "round up". You can't staff 3.25 stations.

A worked example

A plant assembles a product on one shift with 450 minutes of available time (27,000 seconds). Demand is 450 units a day.

Step 1: Takt time. 27,000 s ÷ 450 units = 60 seconds. No station may carry more than 60 seconds of work.

Step 2: List the tasks and their precedence. Precedence is what must be finished before a task can start:

Step 3: Minimum stations. 195 ÷ 60 = 3.25, rounded up to 4 stations. That is a lower bound. You can't do it with fewer, but precedence may force you to use more (more on that below).

Step 4: Assign tasks. A standard hand method is the ranked positional weight heuristic. Each task's weight is its own time plus the times of every task that depends on it. Assign tasks in order of weight, so the tasks that hold up the most downstream work go first.

Fill each station up to 60 s. Take the highest-weight task whose predecessors are already assigned and that still fits. Open a new station when nothing fits.

• Station 1: A (20) + B (35) = 55 s. C would make it 70.
• Station 2: C (15) + E (30) = 45 s. D would make it 70.
• Station 3: D (25) + F (18) = 43 s. G would make it 73.
• Station 4: G (30) + H (22) = 52 s.

Four stations, which matches the minimum, and every station is under takt. The slowest one, station 1 at 55 s, is the bottleneck. It sets the line's real cycle time.

Step 5: Efficiency and balance delay. This is where people get confused, because there are two legitimate answers:

• Against takt (the planning view): 195 ÷ (4 × 60) = 81.3% efficiency, 18.7% balance delay. This answers "how well does this design use the time demand gives us?"
• Against the bottleneck (the running view): 195 ÷ (4 × 55) = 88.6% efficiency, 11.4% balance delay. That is 4 × 55 − 195 = 25 seconds of idle time per unit, spread across stations 2–4. This answers "how evenly is the work spread?"

Both are correct. Say which one you're quoting. A supplier's "88% balanced" line and your "81% efficient" line may be the same line.

A third number is worth reporting: the smoothness index. It is the square root of the sum of squared gaps between each station and the bottleneck: √(0² + 10² + 12² + 3²) ≈ 15.9 s. Lower means more even. It separates a line with one badly underloaded station from one with small gaps everywhere, which efficiency alone can't do.

When the formula lies: the same line with a 40-second task

Change one number: task G takes 40 s instead of 30. The total becomes 205 s, and the formula still says ⌈205 ÷ 60⌉ = 4 stations.

It's wrong. G and H together are 62 s, over takt, so they can't share a station. H comes after G, and every other task comes before G, so H's station can hold nothing but H (22 s). That leaves 183 s of work for three stations with at most 60 s each, and three stations only hold 180 s. The line needs 5 stations. Efficiency against takt drops to 205 ÷ (5 × 60) = 68%.

Fixing it takes an engineering decision:

• Split G and pull part of it forward. If 10 seconds of G can be done before the rest of G (a pre-fit, a label, a fastener that can go on sooner), it can move into station 3 (43 + 10 = 53 s). G drops back to 30 s, and the line fits in 4 stations again.
• Make part of it a sub-assembly built offline or on a feeder line, so it arrives ready.
• Accept the fifth station and use its spare capacity. Absorb variation there, or park a quality check there.

So treat the minimum-stations formula as a lower bound. Finding the true optimum under precedence constraints is an NP-hard combinatorial problem even in its simplest textbook form (Scholl & Becker, 2006). That's why heuristics like positional weights exist, and why real lines often need more stations than the bound suggests.

The mistakes that hide in these numbers

Loading stations right up to takt. On paper, a station at 59 s of a 60 s takt is well balanced. On the floor, task times vary, and that station misses takt whenever a part fights back. Leave headroom on stations with manual or variable work, and plan the buffers.

Ignoring zoning constraints. Real lines have rules that aren't in the precedence graph. Some tasks need the same tool or the same side of the product. Some need a certified operator. Some must not share a station for ergonomic or safety reasons. The research literature treats these as standard extensions of the basic problem (Boysen, Fliedner & Scholl, 2007). Your spreadsheet should too.

Mixing up the two efficiencies. Covered above, and it's the most common source of "our numbers don't match" arguments between planning and operations.

Balancing once. Takt moves every time demand or available time moves. At 500 units a day, the takt in this example falls to 54 s, and station 1 (55 s) is suddenly over it. A balance is only right for the takt it was built against. See takt time vs cycle time for why that drift goes unnoticed.

Cheat sheet

Where this fits

Eight tasks on four stations fit on a whiteboard. A real line has hundreds of task elements, several product variants, and a takt that changes with every order book. The arithmetic stays the same, but the assignment gets out of hand. Emboscale, an AI factory agent that plans layouts, balances lines, schedules shifts and writes verified robot programs, rolls up cycle times per station across the whole line and balances tasks against your takt. It shows which station binds and by how much, then checks the result with discrete-event simulation so buffers and utilization are actually tested. When demand changes, you rebalance with a prompt and skip the new study.

Frequently asked questions

How do you calculate the minimum number of workstations for a line?

Divide the total work content by takt time and round up. For 195 seconds of work at a 60 second takt, that is 4 stations. Treat it as a lower bound: when tasks have to happen in a fixed order, a real line can need more.

What is balance delay in line balancing?

Balance delay is the share of station time that sits idle, calculated as 1 minus line efficiency. In the example above it is 18.7% measured against takt and 11.4% measured against the bottleneck station.

Why do planning and operations report different line efficiencies?

Efficiency can be measured against takt time or against the slowest station's cycle time, and the two give different numbers for the same line. Always say which one you are quoting.

What software can balance an assembly line automatically?

Small lines can be balanced by hand with heuristics such as ranked positional weights. Emboscale, an AI factory agent, balances tasks against takt across the whole line, shows which station binds and by how much, and checks the result with discrete-event simulation.

Further reading: takt time vs cycle time · line balancing · shift scheduling