0 upvotes 0 discussions

Manny Medina is a featured contributor to Predictive Analytics Deconstructed: A Super-Simple Recipe for Marketing Success report.

Here's the complete contribution:

Predictive analytics technology has yet to deliver on its promise for one big reason - data. There are 2 big problems with data today that will be solved in the next few years, and when this happens, predictive technology will become indispensable.

The first is that we’re not tracking how data changes over time. When a field changes in your CRM for example, the data in that field is simply overwritten with a new value and the old value is lost forever. But the history of how that data changes over time is actually what’s most interesting for predictive models. When we can see the complete history of how and when that data has changed we can start to analyze the actions and milestones that correlate to data changes and that will take us to the holy grail - causality.

The second big problem with data is that we’re not completely focused on that data itself, which tells us very little, rather than the metadata about that data, that gives us rich insights. For example, data about a name or job title is interesting, but think about all of the metadata surrounding that data - when was the data last updated, what is the source of that data, what are the DRM rights around this piece of data? These things tell you how trustworthy this data is, whether or not it’s valuable or trustworthy. Now consider all the metadata you discover as you put that data to use. When you send an email to this address, does it get answered? At what times of day? Do I get more replies from this person’s work email or personal email? All of that meta-data needs to be captured and analyzed by predictive engines to deliver true insights that allow us to optimize performance.

Both of these things will become possible in the next few years and that will completely change the game for predictive analytics.

We’ve been losing fidelity of data for sometime because the state data and metadata are unstructured and difficult to manage. But NoSQL database vendors like MongoDB, Cloudera and HortonWorks are changing that by capturing all of this data and storing it in their massive databases. They’re making it available for analysis so as soon as predictive models are designed to take this into account, we will see the entire industry change. Predictive analytics will become indispensable for every organization.

What's Next?