Anybody who has attempted or even researched a Machine Learning and Artificial Intelligence project is familiar with the adage of “garbage in, garbage out“. Put another way, if the data that you feed into your analysis is flawed then your results will be too. While this problem isn’t new it is still a constraint that limits pilot projects from scaling to enterprise transformations. A large and constraining element of garbage in, garbage out is that your datasets have some truth to them, or treasure to follow the analogy. But nobody has the time to filter garbage out. You might not even have the mechanisms to distinguish trash from treasure.

My kids have a fun mechanism: a toy Caterpillar backhoe. They love using it to scoop pebbles and sand to transform a presentable, water-wise yard into an uneven landscape that represents an imaginary, magic-filled land. The sediment has some value in making mini-mountains, but every once in a while they unearth something with a bit of shine. It could be a quarter, or a lost seashell they buried weeks before. And sometimes they find screws. My favorite screws represent a time that I was momentarily more focused on the structure I was assembling with them and didn’t notice the reach of their curious hands; these might help with strengthening a structure that’s currently missing a fastener. Least favorite are rusty ones unintentionally left behind by a previous occupant, of no use to me and dangerous to them.

Business data is similar. Viewed as deposited information, data might be nice on the eyes but offers few insights. When tools and hands get dirty data can represent a story. And every once in a while we come across something that is more interesting than the rest. Yet, much to my kids’ dismay, spending all day playing with it simply isn’t practical. How do we craft stories and find the interesting bits when we’re called to other activities as the data continues to pile up?

One’s Trash is Another’s Treasure

One way is to disrupt the garbage in, garbage out problem by extorting value out of what is traditionally viewed as trash. In the playtime example, when my kids find those rusty screws – trash, to be sure – we’re all better off. I thank them for finding the rusty screw. They’re not old enough to understand oxidation, but something that changes colors is cool! It’s disappointing that we can’t use it, but the kid who found the rusty screw feels good about making the yard safer for his brothers. We LEARN from the trash. Unfortunately for all of us, I can’t let them dig the yard up 24/7.

The same approach can be applied to business data. A technology called process mining delivers the potential to learn from anything. Some of the most interesting insights come from data that would have otherwise lost its utility long ago. Process mining gleans from what an information system processed, and when. Pieced together these forgotten data points tell us a story about the sequence of events – the process. And not just one iteration. Process mining ingests every instance that the process was repeated. Collectively these process flows tell a powerful story about how time, resources, and money are being spent. With fast verification of how the process is supposed to behave, every variant can be categorized into compliant vs non-compliant buckets. Businesses know there are non-compliances, but observers lean forward with intrigue the first time they see the most frequent inefficiencies pulse red while units float along these wasteful paths in accelerated time-lapse mode.

One of the most “boring” process mining summaries I’ve seen still provided a level of insight that would have been hard to prove without our solution. The manufacturing process being modeled seemed almost perfect. What’s wrong with that? I’ll state the obvious – even the best manufacturers would hesitate to call their processes perfect. This company had plenty of data points to prove that their manufacturing had occasional issues. We discovered that what was going into their system didn’t reflect what was happening on the floor. It’s easy to believe that along the way somebody was rewarded for making the data look good (or reprimanded for it looking bad), despite reality. This manufacturer learned that they needed to refocus on enabling good production reporting before embarking on the planned data automation project. A moderate amount of disappointment in the present was more than a fair trade for preventing a retrospective with missing ROI. Even though the historical data failed to represent reality, process mining leveraged the “garbage” to save future time and money.

Finding Treasure in the Trash with Process Intelligence

Manufacturing and supply chain enterprises find treasure in the trash with QAD Process Intelligence. They also solve the 24/7 challenge by integrating Process Intelligence to their business systems like ERP, EAM, logistics and transportation execution, manufacturing execution and supplier relationship management. What Process Intelligence may lack in Big-4 consulting partnerships and marketing, QAD more than makes up for with pre-packaged use cases paired with industry and operational domain expertise that delivers fast ROI in a scalable model.