Menu
Book a demo
Adding a New Model to a Running Assembly Line: Costs and Checks

Adding a New Model to a Running Assembly Line: Costs and Checks

What a new model costs the products already on the line, the levelling rule that absorbs overcycle, and a launch checklist. Case: Toyota Indiana's KEI-01.
Adding a New Model to a Running Assembly Line: Costs and Checks

In short: Adding a new model to a running assembly line costs the models already on it: trial builds take line time, stations that cannot fit the new work content overcycle, operators learn on live product, and a low-volume model tempts the plant to batch it. A study of North American car plants measured an average plant-level productivity loss of 12% to 15% per product launch. The losses shrink with a station-by-station balance check, a levelled sequence and training before start of production.

  • Case: on 10 November 2026 Toyota announced USD 34 million and about 50 jobs to assemble the KEI-01, an intentionally low-volume, right-hand drive, 25 mph utility truck, at its Indiana plant from the second half of 2027. The plant already builds four models with nearly 8,000 people.
  • A new model that runs 30 seconds over takt at one station is absorbed if it comes every 10th unit and the other models take 116.7 seconds or less there. Built in batches of 30, it stops the line 120 minutes a week.
  • Example plant: batching costs USD 121,720 to USD 129,720 a year in idle line labour; a 12% to 15% launch loss over a 12-week window costs USD 140,221 to USD 186,797.

Guide, updated 16 November 2026. Sources are listed at the end.

When a new model joins a line that already builds two, three or four products at a steady takt, the people who run that line keep their targets while the newcomer is learned. That is where launch cost hides: not in the capital budget, but in the output the running models lose for weeks. This guide is about that cost from the plant floor: how to see it before start of production, how to sequence and balance the new model in, and how to measure the launch with OEE by model so the running products are protected. It sits next to a new line ramp-up plan, which covers a line built from scratch, and it uses the same line data as OEE software.

What a new model costs the models already running

The best measured number comes from car plants. Gopal, Goyal, Netessine and Reindorp, writing in Management Science in 2013 with plant-level data from North American automotive plants for 1999 to 2007, found that a product launch brings an average productivity loss of 12% to 15% at the plant level, about USD 42 million to USD 53 million per launch. Three things reduced the loss: product mix flexibility in the body shop, the plant's past launch experience, and past experience building similar products on the same platform. Platform experience lasted; launch experience at the same plant faded faster. A flexible plant with the right platform experience could recover about USD 31 million per launch.

That is a car plant number, but the mechanism travels to any assembly line that adds a model. The loss comes from five places.

  1. Trial builds. Pilot units go down the live line, slower than takt, with engineers at the stations.
  2. Work content that does not fit. The new model needs different or extra work at some stations. Where that work exceeds takt, the station overcycles.
  3. Sequence. A low-volume model is tempting to batch "to get it out of the way", which turns a small overcycle into a line stop.
  4. Learning on live product. Operators build an unfamiliar sequence while the familiar ones keep coming.
  5. Parts and logistics. New part numbers at the line side, new kits, new racks and a new end-of-line test.
Batch the new model, or level it?Line stop minutes a week, Example plantC batched, 30 at a time120 minC every 10th unit0 minC: 30 s over a 120 s takt at one station12-15%productivity lossper car launch$42-53Mper launch,study average116.7 smax A/B timeto absorb CLevel the new model in;do not batch it in.
Line stop minutes a week when a new low-volume model is batched or levelled, Example plant, with the launch loss measured in North American car plants. See the source.

Researchers at the University of Zaragoza make the same point for manual lines: Miqueo, Torralba and Yagüe analyse labour productivity losses during model changeovers on mixed-model manual assembly lines as a first step toward more flexible assembly. The changeover from one model to the next is where they look for the loss, and a new model adds new changeovers to each shift.

Case: Toyota Indiana and the KEI-01, November 2026

On 10 November 2026, Toyota introduced the Scion by Toyota brand and its first vehicle, the KEI-01, to be assembled at Toyota Motor Manufacturing Indiana from the second half of 2027. The same day the Indiana Economic Development Corporation announced the investment.

ItemKEI-01 projectToyota Indiana today
InvestmentUSD 34 million: new assembly operations, production and logistics upgradesmore than USD 8.2 billion to date
Peopleabout 50 new jobsnearly 8,000 associates
Productsone ultra-compact utility truckthree Toyota models and one Lexus model
Startsecond half of 202730 years of production in 2026
Volume"intentionally low-volume, niche"; a fleet pilot firstnot stated
Productjust over 11 ft long, under 5 ft wide; 660cc three-cylinder, CVT, selectable 4WD; up to 600 lb payload; right-hand drive; 25 mph top speed; built to FMVSS No. 500not stated

Sources: Toyota and IEDC releases of 10 November 2026. Neither gives a production volume, a line, a shift pattern or parts sourcing.

Set against the plant, the project is small: USD 34 million is about 0.4% of the more than USD 8.2 billion invested in Princeton, and 50 jobs add about 0.6% to nearly 8,000 (our arithmetic). Small is the hard case. The release facts point to four planning questions any plant adding a low-volume model faces.

  • Same line or its own area? The IEDC says the money will establish "new vehicle assembly operations" and upgrade production and logistics infrastructure. Neither release says whether the KEI-01 joins an existing line. A mid-size plant has to make the same call, and it is the first decision in the table below.
  • Different work content. The KEI-01 is right-hand drive, has a 660cc engine and a CVT, and is under 5 feet wide. Each of those changes station work: steering and pedal parts on the other side, a different powertrain marriage, different fixtures. A balance check station by station is the only way to know where it overcycles.
  • A different end-of-line test. The vehicle is built to the federal low-speed vehicle standard. FMVSS No. 500 caps the top speed attainable in 1.6 km at 40 km/h (25 mph) and lists the equipment each vehicle must carry, from headlamps and turn signals to a windshield, seat belts and, for electric or hybrid vehicles, an alert sound. The final test and audit for a new model has to check what its standard asks for, not what the other models' standards ask for.
  • Low volume, high temptation. "Intentionally low-volume" and a fleet pilot mean a small, uneven order book. That is exactly the volume a plant is tempted to batch. The sequence decides whether the newcomer costs minutes or hours.

The sequencing rule that protects running models

The Lean Enterprise Institute defines heijunka as levelling the type and quantity of production over a fixed period, by quantity and by type of item. Toyota's own description of just-in-time says to make products "at the pace at which they're sold". For a new low-volume model on a running line, levelling is not a philosophy point. It is arithmetic.

Take one station. Takt is T seconds. The new model C takes T + d seconds there, d seconds over. If C comes once every n units, the operator has n minus 1 units of the other models to win back d seconds. That works if the other models take no more than:

Maximum time for the other models at that station = T - d / (n - 1)

With a 120-second takt, a 30-second overcycle and C every 10th unit: 120 - 30 / 9 = 116.7 seconds. If the other models already take 118 seconds at that station, the rule fails and the station needs a fix before launch: move work to a neighbour, add a helper for C units, or change the sequence. If they take 112 seconds, the line absorbs C without a stop.

Batching breaks the rule completely. Thirty C units back to back put 30 x 30 = 900 seconds of overcycle on that station, a 15-minute stop for the whole line unless someone covers it. For more on the trade between levelled and grouped sequences, see heijunka vs mixed-model, and for the station maths behind it, line balancing.

One C in every 10 unitsTakt 120 s; C takes 150 s at one stationABABABABACLevelled: 30 s won back over 9 unitsworks if A and B take 116.7 s or lessBatched: 30 C in a row = 900 s overA 15 minute line stop for every batchunless someone covers the station.
The levelling rule for one station: the other models must leave enough slack to absorb the new model's overcycle before it comes round again.

Worked example: a low-volume model on a two-model line

Every operating figure is Example data. The wage anchor is the BLS's.

  • Example plant: a mid-size assembly line building models A and B at a 120-second takt, 30 units an hour, two 8-hour shifts, 5 days: 80 hours and 2,400 units a week, with 40 operators on the line per shift.
  • The new model C: 10% of volume, so 3 units an hour or 240 a week. It runs 30 seconds over takt at one station; everywhere else it fits.
  • Wage anchor: the BLS average hourly earnings of production and nonsupervisory employees were USD 30.43 in manufacturing and USD 32.43 in durable goods in October 2026. These are earnings, not full labour cost, so the money below is a floor.

Step 1: batched. Planning builds C in batches of 30, one batch every 10 hours of running, 8 batches a week. Each batch stops the line for 15 minutes: 8 x 15 = 120 minutes a week, or 60 units of output at 30 an hour. Forty operators idle for 2 hours is 80 labour hours a week: 80 x 30.43 = USD 2,434.40 to 80 x 32.43 = USD 2,594.40, so USD 121,720 to USD 129,720 a year over 50 weeks.

Step 2: levelled. Sequence C as every 10th unit. At its overcycle station, A and B take 112 seconds (Example data), under the 116.7-second limit, so the station recovers before the next C arrives. The line loses no minutes to C's work content. The cost moves to planning: a mixed sequence needs parts for all three models at the line side all day.

Step 3: the launch window. Even levelled, the launch itself costs. Apply the study's 12% to 15% plant-level productivity loss to this line for a 12-week launch window (the window is Example data). The line has 40 x 80 = 3,200 labour hours a week; 12% to 15% is 384 to 480 hours. That is 384 x 30.43 = USD 11,685.12 to 480 x 32.43 = USD 15,566.40 a week, and USD 140,221 to USD 186,797 over 12 weeks. In output it is 288 to 360 units a week the running models do not get.

Step 4: where the launch loss goes. The five sources above decide how much of that range a plant pays. Trial builds in planned gaps instead of mid-shift, a balance check that finds the overcycle station before start of production, training on a mock-up or off-line jig, and kits staged for the mixed sequence all move the plant toward the low end, or below it.

The point of the example is the comparison. Batching a small model is a recurring cost of the order of USD 120,000 a year on this line. The launch loss is a one-off of a similar size. Both are visible only if OEE is tracked by model and by station, so the plant can see which product lost the minutes.

Book a demo and see OEE per line and shift.

Same line, own cell or separate area: deciding where the new model goes

OptionFits whenMain costRisk to running models
Same line, levelled sequenceWork content is close to takt at most stations and the other models leave slack where it is notLine-side parts for all models, sequence planning, trainingLow if the balance rule holds at every station; trial builds still cost line time
Same line, batchedChangeovers between models are long and cannot be cutLine stops at overcycle stations, inventory between batchesHigh: every batch hits the running models
Same line with a side station for the extra workOnly a few stations overcycle and the extra work can be done off the main flowA side station, a helper, handling in and outMedium: the side station becomes a buffer to manage
Its own cell or areaThe model is very different in size, test or standard, or volume is too low and uneven to levelSpace, fixtures, people, a separate end-of-line testLowest on the line; the new area has its own ramp

Questions to settle with engineering and planning before you choose:

  • At which stations does the new model's work content exceed takt, and by how many seconds?
  • What do the existing models take at those stations today, measured, not from the routing?
  • What is the expected order pattern: steady, or lumpy like a fleet pilot?
  • Does the new model need its own end-of-line test or standard, and where will it run?
  • How long are the changeovers between models, and can they be done inside takt?

If the answer is a mix of many models at low volumes, the wider strategy is in the guide to high mix low volume manufacturing. Food and beverage plants face the same decision for a new product on a filling or packing line.

Who does what before the new model's first unit

Plant manager

Make the line-or-area decision with numbers, not instinct: the overcycle stations, the existing models' slack, the order pattern. Protect the running models' targets explicitly during the launch window, or the launch will be judged by the people whose output it takes. Toyota's KEI-01 money pays for new assembly operations and production and logistics upgrades; a mid-size plant should price both options, line and own area, before it defaults to the existing line.

Production manager

Own the sequence. Write the levelling rule into the schedule and refuse batch requests unless a station check shows they are harmless. Track OEE by model and by shift from the first trial build, with stop reasons operators add at the line, so the launch loss can be split into trial builds, overcycle, learning and parts. Keep takt time fixed through the launch and absorb the new work with balance changes, not with a slower line.

CI and engineering manager

Run the station-by-station balance check for the new model against measured times for the existing ones. Fix every station where the levelling rule fails before start of production. Plan trial builds into planned gaps, and set up training on a jig or mock-up. Build the end-of-line test for the new model's own standard, as the KEI-01 must meet FMVSS No. 500 and not the rules for the plant's other vehicles.

This quarter: a new model readiness checklist

  1. List every station, the new model's expected work content there, and the takt.
  2. Measure the existing models' times at the stations where the new model is over takt.
  3. Apply the levelling rule (takt minus overcycle divided by n minus 1) for the planned mix and mark each failing station.
  4. Fix or relieve each failing station: move work, add a helper for new-model units, or a side station.
  5. Decide the sequence rule (every nth unit) and write it into the schedule, with a ban on unplanned batches.
  6. Book trial builds into planned stops or weekend time, with a count of units and hours.
  7. Start OEE by model and shift now, so the launch has a before and after.
  8. Write the new model's end-of-line test from its own standard and specification.

Measuring the launch while it happens

A launch that is measured by model and station gets better week by week; one measured only as a line total hides which product paid for it. The line data needed is plain: good units by model, stops with reasons, and the time each happened.

Fabrico is a cloud OEE and light MES platform. It shows OEE and losses by machine, line and shift, from the PLC or a Fabrico IoT device, with reasons operators add on a tablet or phone. On manual stations, AI cameras (an add-on) count output. More on light MES and OEE with computer vision.

Book a demo and see losses per line and shift.

What people ask next

How much productivity does a new model launch cost a plant?

A Management Science study of North American car plants (Gopal, Goyal, Netessine and Reindorp, 2013) found an average plant-level productivity loss of 12% to 15% per product launch, about USD 42 million to USD 53 million. Mix flexibility, past launch experience and experience with similar products on the same platform reduced it.

Should a low-volume model be built in batches or mixed into the sequence?

Mixed in, if the line can absorb it. A model that runs d seconds over takt at a station is absorbed when it comes every nth unit and the other models take no more than takt minus d divided by n minus 1 there. With a 120-second takt, a 30-second overcycle and one in every 10 units, that limit is 116.7 seconds. Batching puts the whole overcycle on one station at once and stops the line.

How do you check whether a new model fits an existing line?

Station by station: compare the new model's work content with takt, and measure what the existing models take at each station where the new model is over. Where the levelling rule fails, move work, add a helper for the new model's units or a side station before start of production.

When should a new model get its own cell instead of joining the line?

When it is very different in size, test or standard, when its volume is too low and uneven to level, or when too many stations would overcycle. The cost is space, fixtures, people and a separate end-of-line test; the benefit is that the running models are not touched.

How long does a new model launch affect the line?

It depends on how much trial building, overcycle and learning happens on the live line. Plan a launch window, track OEE by model and shift from the first trial build, and treat the end of the window as the point where the running models are back to their pre-launch output.

What should OEE tracking look like during a model launch?

OEE and good units by model and by shift, with stop reasons added at the line, so the launch loss can be split into trial builds, overcycle at specific stations, learning and parts problems. A line total alone hides which product paid for the launch.

Sources

Sources accessed on 16 November 2026 unless marked as accessed on another date. Dates are the publication dates shown on each page.

Track OEE and downtime in real time
See where lines lose time
Book a live demo.
See where lines lose time
See where lines lose time

Real Fabrico tiles, sample data.

By clicking the Accept button, you are giving your consent to the use of cookies when accessing this website and utilizing our services. To learn more about how cookies are used and managed, please refer to our Privacy Policy and Cookies Declaration