Increase Inventory to Increase Performance

Short lead times are a tremendous competitive advantage.  Little’s Law explains that lower work-in-process (WIP) inventories directly correlate with short lead times.  Lean pushes for one-piece flow to drive the lowest WIP possible and to generate the shortest lead times possible – and, therefore, great competitive advantage.  This is the best business solution, right?

Wrong!  First, Lean does not push for one-piece flow everywhere in the plant.  A quick review of the bible of Value Stream Mapping, Learning to See by Rother and Jones, reveals that a Pacemaker operation should be selected and fed from a supermarket.  This supermarket protects the Pacemaker’s performance from being impacted by preceding operations.  A supermarket is a volume of structured inventory – not one-piece flow. So, even Lean says you need a certain amount of inventory in specific locations.

Second, the lower you take your WIP levels, the more you are impacted by operational variation, making output tougher to control and predict. 

·       Total Productive Maintenance (TPM) can help reduce variation due to machine performance – but it can’t eliminate it. 

·       Six Sigma can help reduce many causes of process variation – but it can’t eliminate it. 

·       Additionally, human beings don’t always perform consistently hour to hour or day to day. 

This means that all production operations have inherent variation – variation in quality, quantity per hour, uptime.  Worse still, that variation accumulates through the steps of a production process rather than averaging out.  The result is a whip effect that adversely impacts customer ship dates.

Since we can’t completely stop the variation, and since we struggle to control it, how do we keep output of the plant consistent so we can meet customer expectations and still make a profit without carrying excessive inventory?

Earlier we talked here about how bottlenecks constantly moving around a plant creates chaos.  We talked about the need to decide where your constraint should be and how you can use various tools to help drive the bottleneck to the constraint. 

Once you have accomplished that, you will need to protect the constraint from the variation accumulating throughout the processes that feed it.  This is critical since, by definition, the constraint controls the throughput of the entire plant.  It defines the capacity of the entire facility.  If the constraint stops running for an hour, that is an hour of capacity lost for the entire facility!  Having extra inventory in front of the constraint helps ensure it doesn’t stop due to lack of parts.  It gives you time to ‘catch up’ if a machine in the flow feeding the constraint breaks down.  It also gives you time to inspect those parts to make sure you don’t waste constraint time processing a part that is already defective. 

Since this inventory buffers the constraint from everything that happens prior to it, let’s call it a buffer.

How large should the buffer be?  You can undertake an extensive statistical analysis, or you can make a rough guesstimate and adjust as needed.  In most operations, 1.5 to 2 days of buffer stock is a good place to start.  Monitor the buffer levels over time (track the data!).  Record the issues that cause the buffer to increase and the ones that cause it to decrease.  Once you have a good feel for how feeding processes impact the supply of parts to the constraint, start lowering inventory between the feeding processes.  As those WIP levels begin to stabilize, start lowering the size of the buffer.  You will probably end up stabilizing the buffer at somewhere around ½ day of inventory. 

Some folks won’t be comfortable with this S.W.A.G. approach but, in the long run, it will save a lot of time and arguments.  The reason is, if you conduct a statistical analysis, people will argue about various aspects of the analysis.  You can easily end up with Analysis Paralysis and get nothing done.  If you do happen to get your analysis approved, it will be implemented, you will monitor performance of the buffer, and then you will adjust it accordingly.  My recommendation just saves you the analysis and the subsequent arguing.

So far in this series, we’ve talked about bottlenecks, constraints, and buffers.  There are a few additional tools needed to make this system work smoothly for you.  One of them is using the proper measurements.  Look for that in our next article.

Our goal for all clients is to have true predictability and control of their business.  This means that, with a reasonable backlog of orders, somewhere about the middle of a given month, you can accurately predict the bottom line of your Profit and Loss statement for that month!  In other words, you can predict what your accountant is going to tell you weeks before they tell you, because performance of your plant is so predictable.  The business is also much easier to manage.  Frequently, we have clients thank us for giving them their lives back.

If you would like to have this level of predictability and control for your plant, drop me a line at [email protected] .