Cracking the Code: Lora Cecere, Supply Chain Insights

Cracking the code on unstructured data is a tough but important nut to crack. It is a key competency that will define the next generation of consumer products leaders. Today’s architectures are designed to answer “the questions that we know to ask”. It takes many forms: Reporting, alerts, scorecards and dashboards. Unfortunately, most companies are so focused on trying to use transactional data that they have little time to think about the “art of the possible” where structured and unstructured data come together to answer new questions. As companies define these new architectures, they will be able to “listen, test and learn”. To do this, the data needs to be “freed to answer the questions that we do not know to ask”.

In consumer products, there are two initiatives that are driving this realization: Digital path to purchase and safe and secure supply chains (one example of this is food safety).

Digital path to purchase initiatives are attempting to influence the shopper at the “moments of truth” in the decision cycle to buy products. It starts with trying to influence the shopper to put the item on the list, and continues to the store to influence the shopper to put the item in the basket. It is the confluence of mobile, social, e-commerce and geolocation technologies. In the process of implementation, consumer manufacturers will come to the realization that they do not know demand data very well. Syndicated data is too dirty and late, and current demand signal repository architectures give insight on what was bought when, but fails to tell us why. As companies attempt to build digital path to purchase initiatives, the use of unstructured data will grow in importance. In these initiatives, it will be our ability to “listen” that will define success. We will not be able to test and learn without the ability to use semi-structured data well.

In food safety initiatives, the discussion is similar. The use of sentiment data from social media and user-based comments can give us an early warning on sensing a problem or a potential problem. The ability to use this data to listen can take two to four days out of the problem identification cycle. We cannot know that there is a problem unless we design architectures to listen for what we do not know to ask.

When customers finally “tell us” through conventional means, like call centers and e-mail, it may be too late. This will require the deployment of natural language processing systems based on rules-based ontologies that will allow us to listen before we learn. This data will then need to be combined with structured data to trace problems and track root cause. With the outsourced supply chain the data will come from many sources requiring the harmonization on the supplier data.
 
I believe that this next generation of business intelligence offers many, many possibilities, but it also requires a different mind set. Instead of controlling data, we need to free it and employ learning systems to use the data in multiple forms. In the process, traditional master data approaches will be seen as archaic as the dinosaur.

I had a conversation the other day with Amazon.com, and asked them about master data management. They laughed. The reason? It was because they realized a long time ago that the traditional methods of master data management were not up to the task. Instead, they use semi-structured data forms that can be tagged and indexed and used on demand.