Dear Denis, you need to postpone the final assembly stage of the customised products starting from common sub-assemblies that can be configured in different ways. The inventory should so kept at the subassembly level. Depending onto your delivery lead time you can "postpone" final assembly to lower levels working on the sub assemblies of sub assemblies and so on. Even the forecast activity can be done at lower manufacturing "bill of material" level than the final product.

So the choice is strictly dependent on the market you have to fulfil and the product you're realising.

Hope this could help you in proceeding your research.

Sincerely

Flavio

hi Dennis...

I like the question and I too like what Flavio suggested to you to postpone the assembly stage with the inventories at sub-assembly level. The good thing is that it concerned only with unique products. You may apply JIT to monitor the inventory closely too. However, I would cautious you that this depending on the pull factor of quantities the market demand. Usually for pull market the quantities are with large amount and that means more work need to be done and monitored including the space saving for raw materials and numbers of production lines required. That means the man power too need to be increased.

Hope this could some ideas about it. Cheers.

Dear Denis,

I agree with Flavio. In that case you would be implementing a Conwip-Assembly. Also according to Adam.

The pull type control environments with multiple products can be done in two ways (M-CLOSE and S-CLOSE loops). One considering the global inventory level and calculating the number of cards according to the average customer demand. Or it can be done considering the inventory and pull loop independently for each product type. This is not feasible for environments with many types of products. Adjusting the number of cards is more complicated and troublesome management.

There is a third possibility and manage the system with a "Cobacabana" system ... (Martin Land) managed by cards, focusing on the load of the machines (stations) instead of being focused to manage the inventory system ... It is not exactly pull, but is more suitable for multiproduct environments.

Regards!

Thank you Dear Flavio , Adam and Pedro. Your answers are very helpful, thank you for providing such thoughts.  Flavio, one question. In order to keep inventories at sub-assembly level as you suggested, do you agree that products at least should sustain minimal similarities along them (material , design, facilities). Otherwise it does not work, right? Pedro, thank you also for your answer. Large number of products makes cards management really cumbersome. Adam, thank you for cautions in the work force.

I have had an opinion that Lean Manufacturing and JIT,  were not designed to handle highly customized multi-products or projects environment. The lead-time of an order is not only dependent on its own tasks time; it is dependent on other tasks from different orders being processed, or are waiting to be processed. The assessment is tough because the range of services, products provided is very high, workload is uneven due to the wide range of skills and manufacturing process levels required to complete competing orders.

I really appreciate your answers and I will investigate each recommendation of you. cheers

In semiconductors, we went from 90 day cycle time to less than 30 days within one year just by reducing inventory and focusing on improving bottleneck cycle time.  For example, we could produce 20k wafers per month - so... you don't want your inventory to exceed that level.  Everything is moving all the time.  We also had some products that had the same front end design, but were differentiated at the middle.  We would run product and hold them until we received an order - instead of 30 days, it took about 14 days.  If you haven't read "Then Goal" I suggest you do - all our engineers were required to read it - the results were transformative.

Hello Mathias, thank you for your answer.  I have read the book that you mentioned "The Goal" By Eliyahu Goldratt. However, correct if I am wrong,  this books is related the TOC and somehow it deals with a much more stable environment than I have mentioned. Bottlenecks must be identified a priori to get TOC working well. 

My question is regarding to environment such as:

What if we don't know anything what we are going to perform, produce or delivery? Each order is unique in all sense. How is it possible to pull the production in such environments? Pull the production requires a minimal level of intermediate inventory (work in process - WIP) . Assume that every order is different, it requires completely different materials (BOM), different pool of resources and the design is different. How is it possible to pull the production? Why should we keep inventory that situation? thus, if we don't have inventory, how to pull?

Sometimes through the literature , we have seen some misunderstanding, people trying to use some techniques where by nature is not possible. 

I can go further and expand that:

How to dedicate resources and design manufacturing cells for groups of products and pull the production if the services often do not repeat, if they are high customized and there are a wide variety of services? ( JIT and Lean Manufacturing adheres well?)

How to identify a priori capacity bottlenecks if there is no regular work flow? (TOC fits well?);

How to identify critical paths in which multi-projects are running simultaneously if  tasks are prone to large waiting times in resource queues that depend on the current services? (PERT/CPM)

Dear Matthias, thank you again for your answer, I appreciate that.

Denis,

Flavio suggested that you "customise" as late as possible in the process. I worked in a specialised packaging company some years ago where we provided custom packaging on a 24 hour turnaround by making the products from generic components and adding the custom features (printed elements, specific inserts etc.) when order where placed. These custom elements all had SMED style set-ups.

The second approach is to look at the underlying process, rather than the products themselves. Any business providing custom solutions will most likely have one or a small number of processes. If you focus on these, you can improve your flexibility. I worked for a company that custom built all its products - when we introduced JIT, we focused on standardising the process, with good success.

The final method we used was to design our products so that they became generic. We did this with a range of percussion cases. The initial products when we entered the market were different sizes for each distributor. Using an industrial designer, we created a standard range that had 50% fewer cases, but catered for any size/marque of instrument - the distributors loved it, because they could stock a smaller range and meet all their customer's needs - win:win for all parties.

Hope that helps you with your quest.

F

Hi Frank, Thank you for answering my question. I appreciate that.

Regards "Any business providing custom solutions will most likely have one or a small number of processes", I don't know if I agree with that . You are partially right.

Now Imagine shipbuilding , ship repairs, constructions,  manufacturing environments to produce equipment for coal industry, power plant, so on so forth. They have several processes.

Overall I got your point and it makes a lot of sense. Again, I appreciate your thoughts and your time.

have a nice weekend.

best regards,

Denis

Denis,

Glad you got some use from the response. Regarding processes, I suppose what I'm trying to say is that you can move from the complexity of multiple products and services to a more abstract level that shows commonalities. In one project involving 1000+ products, each with a different process flow, I used abstraction to create an inventory control system based on a common measure of priority (days of inventory). The process complexity was reduced by identifying a couple of control points in each process flow that were common across 90%+ of the flows. The system highlighted a) where inventory production was required (relative shortage compared to consumption rate) and b) where excess inventory existed; based on low relative inventory turns. These relative measures were much more effective than static inventory targets, especially when demand for specific products was quite variable.