In a previous post (Starting from Scratch), I outlined significant differences between our Supply Chain Index and existing systems for measuring and ranking supply chains. In this post, I’d like to continue that discussion and outline some of the decisions and interesting insights we have discovered in our ongoing work in building the Index.
Choosing the Metrics
Over the past year plus, we have built a financial database of over 50 metrics for most publicly traded companies around the world reporting annual results. This database is big, to say the least, and includes many metrics which are not necessarily closely tied to supply chain. Thus, the first step in our research became clear: we needed to identify the short list of meaningful supply chain metrics. We did this in three cleaning sweeps.
Initial analysis showed that when a large company dominated a peer group, the result was that that company would “take over” the peer group and the result would be a high ranking for them due to their size, irrespective of their actual supply chain performance. In short, their size manipulated the rankings. So, in order to compare fairly across company size, we removed all raw number metrics (revenue, cost of goods sold, employee count, etc).
We also removed duplicate metrics. The most powerful example is the variety of inventory metrics available. Our database included more than five inventory metrics including
* days of inventory
* inventory turns
* days of finished goods
* days of work in progress
* days of raw materials
* and others
In order to only count inventory performance once, we cut all duplicates and only allowed days of inventory in our short list.
Finally, we omitted metrics which were not dependably available in the database. Employee growth—which measures the change in number of employees on an annual basis—is interesting, but rarely available and thus not reliable enough to use in building our equation.
In total, we cut 40 metrics from the starting point. We ended with 14 supply chain metrics on the short list as shown in the table below.
Linear & Nonlinear
Our first trials focused on defining linear relationships between market capitalization (y) and the list of independent variable metrics shown above. We quickly realized that although linear modeling may be fine in contrived problem sets, it loses some of its power in the chaos of the real world.
We readjusted our approach, and allowed for the possibility that the variables might be related in nonlinear ways, and ended up with a much stronger equation (higher r squared) and more comfortable with the results.
Searching for Relationships
Finally, the journey has illuminated several interesting results (already!) and the discoveries continue. The members of the team overwhelmingly expected a strong relationship (linear or nonlinear) between year-over-year revenue growth and market capitalization performance. The chart of this relationship is shown here for our chemical peer group.
In short, there seems to be no relationship, and definitely no strong relationship, between the two metrics. Over and over again, this work challenges our assumptions about what makes a good or great supply chain and brings data and objectivity back to the table.
Tune in for our upcoming webinar on April 25 for more information on the Index, the steps we took in building it, and a discussion about equations and rankings for several key industries. Here is the link to register. I hope to see you there.