1
$\begingroup$

I am in a factory with many manual (non-machine) operations. Process time variability for most operations is high, leading to highly variable cycle times at the part level. There are efforts to reduce process time and variability, but the current state is highly variable and will likely always be at least higher variability than a machine-based operation. Scheduling is in a bad state and items are frequently late to internal plans.

Orders are entered in SAP based on their due dates, then subassemblies are planned backwards from there based on planned cycle times stored in SAP. Setting those planned cycle times is where I'm looking for the best approach.

The factory has a historic practice of setting planned cycle times in SAP to recent average cycle times. This approach doesn't seem like a good one - wouldn't you be planning to be late roughly half the time if SAP is planning orders to your average cycle time?

My leading thought is that aiming for a service level such as 95% on-time delivery for the deliverable part would lead to very high service level requirements for the subassemblies, meaning planned cycle times would be set very high in comparison to usual practice so that the parts have a high chance of coming out on time. This would cause WIP levels to rise and more space to be used in the factory. However, on-time delivery should improve.

Looking at one example, a part may come out in 2 days or up to 20 with high variability over each part completion. To meet service levels, I may plan in SAP for this part to take, say, 18 days, which would usually not be needed but would be a level which would give us higher confidence in finishing on time.

Additionally, is it better practice to use data on recent history or try to forecast cycle time instead? There are so many factors that can inflate cycle time - vendor shortages, defect trends, poor performance, etc. that feel unpredictable.

Curious if anyone has some good practices for planning cycle times in manual labor, high variability factories.

$\endgroup$
5
  • $\begingroup$ Good question. Forecasting with the average without consideration of variability has limited use, can be misleading. One possible analysis is to see if the variability is between individual assemblies, with a possibility to "smooth it out" within a batch containing a larger number assemblies. If the bulk of the variability comes at the level of an entire batch, there is not much choice other than to reserve the necessary time for a worst-case, or 95% case, or something like that. I would also maybe separate internal vs external. Parts purchasing delays we are all seeing are their own thing... $\endgroup$
    – Pete W
    Nov 8 '21 at 17:54
  • $\begingroup$ Also golden rule of engineering communication in business -- whenever you say "A +/- B", no matter what you write down and no matter how many times you repeat it, management will sooner or later make the "B" disappear. If it is truly important, express it in terms of worst-case A. $\endgroup$
    – Pete W
    Nov 8 '21 at 17:56
  • $\begingroup$ Batching is a bit difficult due to the parts being fairly large and complex but that is an interesting thought. The internal vs. external comment led me to think about another issue - what to do, if anything, with planned cycle time if I have existing stock that is finished awaiting assembly into the next higher part. Does it matter? If I have too much stock, maybe I don't really need to plan the full cycle time. Or maybe SAP can already see and plan that and won't release the lower level order if there is already stock existing. $\endgroup$
    – gtm
    Nov 9 '21 at 11:58
  • $\begingroup$ Re: stocks of intermediate parts/assemblies -- If the yield is anything less than 100%, there will typically be some stock left over. It may also be economical for an upstream process to have a bigger order size. The observation of JIT is that a big stock / buffer can conceal variability upstream of it, and even gives a reason for upstream processes to develop inflexible production strategies. That should NOT be interpreted to mean stocks / buffers ought to be eliminated unconditionally, or eliminated simply on the basis of accounting savings on the depreciation of the inventory... $\endgroup$
    – Pete W
    Nov 9 '21 at 15:07
  • $\begingroup$ ... But in your case, it sounds like you have a significant variability in the production process, and the estimate you are making is the way you communicate this fact into a language that can be understood by the corporate machine. If you want to reduce the variability, you might need to take a deeper dive into the process itself -- not sure if that is within your scope. How the SAP was programmed is important enough that you could try to communicate with whoever knows that, if it affects what you're doing. $\endgroup$
    – Pete W
    Nov 9 '21 at 15:08
1
$\begingroup$

At issue is what are you trying to do with your planning times.

Setting to averages will maximize machine efficiency and plant throughput.

But if you are using those times for Available To Promise (ATP) and order delivery dates, you need a different method. Probably the best method for that is to add wait or buffer time to the order delivery date to account for inefficiencies. This makes the production schedule and the order delivery date possibly a little different. There are manual and electronic methods to escalate orders through the plant according to expected completion time and order delivery dates. Start with The Goal and go from there.

$\endgroup$
1
  • 1
    $\begingroup$ In a way maybe it just is what it is - in a high variability environment, the buffer has to be large enough to handle that variability, causing higher WIP levels but enabling greater on-time delivery. $\endgroup$
    – gtm
    Nov 9 '21 at 11:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.