How to Design for Assembly Line Efficiency

The line is a design constraint. A base that sits without being regrasped, fasteners from one direction, and a test no longer than the mechanical work, are product decisions. They are expensive after the tool is cut. Figures below are hypothetical teaching maths, not BrahmWorks prices, yields, timelines or test results. The documentation a line needs in order to run without its designer is part of How to Take a Hardware Prototype to Production.

Takt time is available production time divided by the units that time must finish. It is not how long your best assembler takes. If the hours and the demand are real, takt is a requirement. Station content fits under it, or you add a line, add hours, or remove work.

Balance is how evenly the work splits. Stations at 20, 22, 21 and 55 seconds are paced by the 55. The others wait. Balance efficiency is the sum of station times divided by the number of stations times the longest. Waiting is often designed in: a cable dress only one person can do, a glue cure nobody timed, a test longer than the mechanical stack. Moving a task is legitimate. Hoping the long station "speeds up with practice" is not a plan.

A regional cable, a different front or an optional sensor is a branch in the work. The station still needs the parts and the decision. Make the option a module at one station, not a different sequence at every station.

A part should arrive in the orientation it is fitted, or the station spends takt turning it. Kit when the line is short and the product varies. Use line-side bins when the part is common and hard to pick wrongly. Left and right parts that look alike defeat both. Make them identical or impossible to swap.

Fixtures hold what the operator used to hold. They are not a pardon for mid-air assembly. A housing that cannot sit until the last screw is regrasped at every earlier station. Each regrasp costs time and risks a loose board. Give the product a stable base early and add parts from above. If a nest is unavoidable, time it and build it. Do not discover it as tape.

Cure times set a minimum the operator cannot shrink. If a cure is longer than takt, move it off the station: a buffer, a travelling fixture, or a joint that needs no cure. Found on the line, it becomes a table of drying units in the aisle. Found in the design, the joint can still change.

A final test that takes four times the takt cannot sit on a single seat in the line without four seats of test, or a design of test that is faster. Long functional scripts, manual button pressing, and a calibration that needs a skilled operator are the usual sources. Split the script. Automate the stimulus. Move a long burn-in off the paced line and give it a clear entry and exit so units are not lost.

In-station checks should catch the defect while the station can still fix it. A polarity check after the housing is screwed shut is a rework path, not a check. The line is efficient when defects are found at the station that created them, and when the repair does not require a private expert. If only one person can diagnose a failure, the line's capacity is that person.

The teaching product is a small countertop device. It is not a BrahmWorks product. Five stations are planned: base and board, cable, top housing, label, functional test. Illustrated times, not a BrahmWorks study, are 40, 70, 45, 30 and 65 seconds. The cable station is long because the lead dresses around a boss the operator cannot see, and because an unpolarised connector is sometimes reversed and then fixed.

Teaching changes, not a measured new balance. A channel in the base holds the lead in one path, visible before the top goes on. The connector becomes polarised so the reverse-and-repair disappears. The functional test, at 65 seconds, is split: a 20-second power and firmware check stays on the line, and a longer radio check moves to a two-seat offline rack so the paced line is not waiting on the air interface. The label moves onto the cable station's neighbour only after the cable time falls, otherwise you have shifted a queue rather than removed it.

The design changes are modest. The line effect is that the longest remaining paced station is the one you plan labour against, not the cable habit.

Hypothetical teaching maths. Not a factory measurement and not a BrahmWorks result.

Assume 7.5 productive hours in a shift, which is 7.5 × 3600 = 27,000 seconds. Assume demand of 400 units in that shift. Takt = 27,000 / 400 = 67.5 seconds per unit.

The illustrated station times are 40, 70, 45, 30 and 65. Sum = 250 seconds. Longest = 70. Balance = 250 / (5 × 70) = 250 / 350 = 0.71, or 71%. The line cannot meet a 67.5 second takt while a station takes 70, even before any stoppage. You are short by 70 − 67.5 = 2.5 seconds on that station, which is enough to miss the day, and the other stations have spare time the customer does not pay you to idle.

After the teaching change, assume the cable station falls to 45 seconds and the on-line test is 20, and the offline radio check is not on this takt. New paced times might be 40, 45, 45, 30 and 20. Sum = 180. Longest = 45. Balance = 180 / (5 × 45) = 180 / 225 = 0.80. Takt of 67.5 now covers the longest station with room. The room is not a victory to fill with extra fiddling. It is where variation, a dropped screw and a question will sit. These times are assumed to make the arithmetic readable. Replace them with your own study before you staff a line. Do not quote 71% as a result.

• Takt is computed from real hours and real demand.
• Every station's content fits under takt, including the check done there.
• The longest station is known, and it is not "whoever is good at cables".
• The product has a stable base early. Later parts stack from above.
• Regrasps and mid-air holds are designed out or fixtured and timed.
• Left and right, or other lookalikes, cannot be binned side by side.
• Cure and glue times either fit takt or leave the paced line on purpose.
• A defect is detectable at the station that caused it.
• A long test is parallel or offline, with a record so units do not vanish.
• Options are a module at one station, not a new sequence at all of them.

No. A fast assembler shows what one person can do, often by skipping hesitations a new hire will hit. Design for a trained operator who is not your specialist. If the time exists only when a particular person is present, the process is that person.

Yes when the test is short enough to fit takt and when it catches what the previous stations just did. No when the script is many times takt. In that case take a short go/no-go on the line and move the long test offline, with serial numbers so a failed unit cannot be packed. A test station that always passes is not balance. It is a walk.

They can hold an awkward product, and sometimes they should, because the use case forces the shape. They do not remove the time to load them. A fixture with three clamps and a rotation can cost more than the screw it was meant to help. Prefer a product that sits. Use a fixture for what remains, and time the load as part of the station, not as free.

Then do not invent one. A cell, where one or two people build the whole unit, can be the efficient process at low volume. The design rules that still apply are a stable base, no lookalike parts, and a test with a limit. The rules you can relax are takt balance across five stations and dedicated fixtures for every step. Build the cell properly rather than a pretend line that waits on one expert.

Bring the station plan, the demand you are staffing for, and the operations that only one person can do. BrahmWorks can look for the content that will pace the line, and for the design change that removes it.

Review the next stage with BrahmWorks