Learning from Machines: What Traditional Manufacturing Can Teach Us About Data Products

As soon as data products rely on continuous incoming data, the framework needs to follow the supply management practices of physical products, otherwise you’ll be forced into reactive management.Looking DeeperGoing back to the physical product analogy, let’s look at a couple real-life scenarios that illustrate why the comparison to physical product is so informative, particularly as relates to supply chain management.For a period, I worked as a buyer for that aerospace company that produced radios..One day, we discovered that there was a bad batch of switching ICs (a type of electronic chip)..Most of the batch failed testing and we needed to buy more to build one of our less popular radios..When I checked with our supplier, I discovered that they no longer sold that chip..Worse, other suppliers had listed it as obsolete..The manufacturer had stopped making them 2 years prior and we didn’t know about it..We had been making the radio for 12 years..We needed a replacement chip ASAP!We were able to do a minor redesign on a sub-assembly to incorporate a newer, smaller chip..That necessitated halting production until the new design was in place, re-testing the product before restarting production and working overtime to backfill the orders to honor customers’ commitments..For a more conservative industry, where the same products are sold for decades, this type of scenario is not uncommon.Photo by Nicolas Thomas on UnsplashAlthough it might not be immediately obvious, this type of situation is actually incredibly similar to what data scientists might encounter..Recently, I was checking data updates for a prototype model..I wanted to verify that all the external contextual data we were using were fresh, so test predictions would be accurate.. More details

Leave a Reply