Want To Be On Time? Better Get There Early!^{©}by Arthur M. Schneiderman We all want to be on time, whether it's to an appointment, delivering our product or service, or in meeting any of our other time based commitments. So what do we do? Estimate how long it will take, subtract that from the due date and that's when we need to start. That's good enough most of the time, but what if we want to meet our commitments virtually all of the time? Our first improvement might be to replace our estimate with actual data. Let's say that we have a product that we produce to order. We've kept track of the historic cycle time for this product and it averages 100 days. We add 10 days to our estimate "to be on the safe side." Place an order with us and we'll quote a lead time of 110 days. But, let's look at the historic data more closely: What you see above is what we call a histogram. In this simulation, I've generated 1000 normally distributed numbers having an average value of 100 and a standard deviation (called s) of 10. The dots show the number of points in this sample having each of the x-axis values. For example, there were 41 points with a value of 99 days. The solid curve represents the cumulative percent with values less than the x-axis value. As you can see, there are no points below 73 or above 131 so the curve touches 0% and 100% at these values. This curve also represents our best estimate, based on past history, of the probability that an individual cycle time will lie below a given value. Now here's where things get interesting. If we make commitments based on the average cycle time, we'll miss them 50% of the time. Even if we add 10 days, we'll still miss our ship date 18% of the time. If we want to be on time 99.8% of the time, we need to add 29 days to our delivery commitment!! What does this all mean?
What is the root cause of this phenomenon? The answer: variation. One of Deming's recurrent themes was the evils caused by process variation. Much of his focus was on variation due to special causes, rather than the common causes described here. Let's look at this same picture after improvement efforts have reduced the average to 75 days and the standard deviation to 5 days.
If our competitors are still operating under an equivalent "old" process, then quoting 100 days will allow us to get our share of the business while being on time virtually 100% of the time. In fact, as you can see above, we can drop our lead time to 90 days with nearly 100% on time delivery. However, if customers value shorter lead times, we may be able to gain market share by quoting significantly shorter lead times at industry standard delivery performance. And customers should value shorter lead times since it reduces (or eliminates) their forecasting horizon and the associated forecasting errors. Or better yet, we can combine improved delivery AND lead time (based on customer's importance weightings of these attributes) into a strong competitive package: shortest lead time, best delivery! Here's another application of this principle:
The number one consumer of executive time is attending meetings, and one of the biggest root causes of time wasted in meetings is late starts. I do realize that in some organizations, getting to a meeting early is interpreted as a sign that you don't have enough to do. Rush in late, give an out of breath apology, and all is forgiven. I leave it to your ingenuity to adjust my advice here to your organization's culture. Of course if you're the boss, here's an opportunity to catalyze a change to a more productive culture. Whether it's wasted time or unwanted inventory, excess variation has costs associated with it. But remember, if you want to be on time, you have to plan to be early, and you usually will be. |

Last modified: August 13, 2006 |