Lean & Assembly Line Throughput Times

Who or what is telling people to look at throughput time (lead-time) through the assembly line when measuring productivity? This number is completely unrelated to assembly line productivity.
I have run into the situation where people charged with doing kaizen on assembly lines (industrial engineers, project managers) are measuring the wrong thing, throughput time. These are trained and educated people. This is happening at name-brand, global manufacturing companies. This week was the third time in two years. Please make this stop.
I don’t have a formal IE background so I don’t know if this is a standard IE practice or if it comes out of a text book. I do know that the definition of “cycle time” in some textbooks is what we call “manufacturing lead-time” rather than the Lean definition which is something like “hand-time” or “operator cycle time”. This can cause confusion at times.
When doing kaizen to design assembly lines for one-piece flow one of the first activities is measuring all of the assembly cycle times (total hands-on time to build one unit, and separately total automatic cycle time). The next step is to calculate takt time for the particular product or product family or value stream. The next step is to balance the workload to takt time and design the layout, work sequence, standard WIP positions, etc (creating Standard Work).
Measuring the time it takes for one unit to make it from the beginning of the assembly line to the end has *no* direct relation to the productivity of the line. While it can be an interesting measure of response time to customer orders, this lead-time number should not be used for designing one-piece flow (or even batch & queue) assembly lines.
After one-piece flow is achieved, only Standard WIP remains in the line so the lead-time calculation is takt time x Standard WIP quantity. If you are doing “fake flow” and pushing, your lead time will be longer and cycle times will be out of balance. However the lead-time of “fake flow” will have no bearing on what Standard Work and one-piece flow should look like.
There is no need to measure this throughput time at all when doing kaizen in assembly since all you are doing is identifying the “wait time” of parts in transit on a belt or sitting on an assembly bench. A quick demonstration has been enough in each case to help people see that throughput time in assembly is irrelevant to productivity.
I have thought about other reasons for this type of measurement but to no avail. I don’t plan on researching it further but if someone knows the source of this practice of measuring assembly line throughput time please let me know.

13 Comments

  1. Manjula

    July 18, 2007 - 4:20 am

    I am doing the vsm, current state of a large volume parental article. How do we calculate cycle time for an activity like mixing, where a solution batch is equivalent to 7 000 bottles?
    On a bottle formation line; 4 bottles are formed by blow fill method, and they move on a conveyor. At the end of the conveyor, semi auto machine are places where 2 bottles each are picked by operators, capped and placed on another conveyor where each bottle is ink jet coded with batch details. These bottles then move into a collection bin, an are then stacked into sterilisation trays manually. How to calculate cycle time here?
    Do I calculate only the value added without the movement on the conveyor or the entire line movement of a bottle? Thanks

  2. Jon Miller

    July 18, 2007 - 8:44 am

    The cycle time for mixing would be measured for one batch. You can then divide this by 7,000 to see if the cycle time is lower than your takt time and balanced with upstream and downstream processes. Show both process cycle time (batch) and time per piece in your data box.
    The same with blow fill. Measure cycle time per cycle of the machine, then divide by four to see whether cycle time per bottle is under takt.
    All cycle time is for cycle “from X to Y” including all work that is done by man or machine, including wasted motion. Conveyance would typically not be part of cycle time but expressed as inventory, which in turn is calculated as part of lead time for the entire line.
    Conveyance one at a time by a person or by a robot arm would be considered as part of the cycle time, however.
    Since the time on the conveyor is in parallel to the manual or machine cycle time, and it is pure conveyance, it is not part of cycle time. If the machine or man had to wait for several minutes of conveyance until a sufficient number of bottles was delivered to begin work, this would be part of the cycle time.
    Be careful also since the definition of cycle time I am using is closer to “process time” or “touch time” as opposed to cycle time which some use for “total time through a process” as in lead-time.

  3. Anonymous

    August 17, 2007 - 2:08 pm

    Hi, maybe it is not exactly about the discussion but I have in mind a question about calculating the manufacturing lead time MLT) which is directly related to the WIP. In theory there are many ways to calculate this value from queuing networks to simulation models. But almost ever of these methods are very time consuming. And I see that most firms say that they have reduced their MLT by some percent. How do they calculate MLT?
    For example think about a factory which has got different transfer batch sizes, different routing possibilities. Furthermore breakdowns of machines and even setup times directly add up to MLT. In addition to these even scheduling extremely affects MLT. So how a single MLT can be calculated practically and in short period of time?

  4. Jon

    August 18, 2007 - 9:54 pm

    The theoretical manufacturing lead times can be calculated as the sum of processing time + delays. Delays would include time in queue, breakdowns of machines, schedule changes, or other non productive activity.
    Where theoretical and practical manufacturing lead times differ, this is an opportunity for kaizen to close the gap.
    Most likely the difference in measured manufacturing lead time before and after is used to calculate percentage reduction, not theoretical.
    Actual manufacturing lead times (practical, not theoretical) can only be measured, not calculated. If the manufacturing lead time for similar products going through the same processes vary too widely across a period of time such as a month, you do not have a stable process.

  5. Anonymous

    September 4, 2007 - 3:16 am

    Hi, I think the problem is how delays are calculated to figure out the theoretical MLT? How companies such as Toyota or Ford calculate these delays to find their theoretical MLT?

  6. SUSHANT

    April 7, 2008 - 3:27 am

    I want to know the difference between TAKT TIME, CYCLE TIME & THROUGHPUT TIME

  7. Alienx

    June 30, 2008 - 1:59 am

    TAKT Time calculated based on forecast provided by customer.
    Cycle time is measured by stopwatch or some other ways, PTS ways,
    Throughput time is the actual output time.

  8. Balaji

    July 30, 2008 - 2:25 am

    I want to know the method of calculating the Manufacturing Throughput time. Also for an entire plant with many manufacturing shops, how to calculate Manufacturing throughput time for entire plant?

  9. Jon Miller

    July 31, 2008 - 2:42 pm

    There are several ways you could calculate the manufacturing throughput time, depending on why you needed to know this number.
    One way is to record the time and attach tags to orders when they are first entered, and then record the time when they exit the overall process. This is not a calculation per se but it is the reality of your throughput time. You will need to do this several times to understand the range and standard deviation, unless your processes are highly repeatable.
    You can also multiply your standard work in process quantity by your takt time. If you don’t have no established standard work in process quantity, your throughput times will vary. If you do not work to takt time, identify the pacing process.
    Another approach is to make a sum of all processing and waiting times.
    Creating a value stream map is a good way to get all of this information in one place.
    The more complex your process, the more work this will take.

  10. Shashwat Sharma

    August 1, 2008 - 3:51 am

    Lead Time, Throughput Time, TAKT Time, Kaizen

  11. rajath

    March 5, 2009 - 4:53 am

    i wanted to know the procedure to calculate the throughput time in a bus body building plant..please let me know soon…i will be very thankfull..also please give some suggestions on how we can reduce the throughput time in a bus body building plant..

  12. navid

    March 24, 2009 - 4:49 pm

    Hi,it would be very appreciated if you answer this question? How delays are calculated to figure out the theoretical MLT?

  13. dhale

    April 21, 2009 - 1:24 pm

    Throughput =
    (Total pieces x 1st process time) + each individual process time + each queue time between processes.
    Not “exact” figure but will give you a good idea of how long it will take to process 10 pieces or 100 pieces through your system.
    This is based on 1 piece flow.