How do we build analytics capabilities?

As we build analytics maturity, we continue to show value through asking questions about the business and showing the impact that marketing is having on the business. Are we engaging our target audience? Are we connecting with our target audience where they are? Are we meeting our audience’s needs?

The first three pillars of an Analytics Maturity curve have these outcomes:

1)      Descriptive Reporting – What happened?

2)      Diagnostic Reporting – Why did it happen?

3)      Predictive Modeling – What could happen?

In the descriptive analytics step, we look at periodic reports that relay sales or marketing engagement. A report shows how many widgets sold last month. A different report shows how many people came to our web site last week. These reports give us a barometer of what happened.

As we procure more data, more specialty data analytics skills in resources, perhaps better quality data, we move into the diagnostic analytics maturity phase. In this phase, we look at the descriptive reports and ask, ‘why did something happen?’ For example, in retail, we look at why did our umbrella sales grow tenfold last weekend. Through another data point, we learn that it rained in that region on Saturday. In this phase, we combine different data sets to learn about our markets and customers.

How do we move to the Predictive Phase of Analytics maturity?

If we can name an action that we want a customer to take, we can predict the likelihood of someone taking that action. In the world of media, we want to engage someone on a banner ad with a given headline message.  As we plan the days of expensive live sales representatives visiting prospects, we want to visit the prospects that are most likely to result in a sale. We make predictions using data and machine learning algorithms. We use historical data to train a model based on attributes of past actions to predict the action will happen in the future given similar conditions. This is predictive modeling. Predictions are modeled on past behavior which means we need high quality data and complete data sets of the features or predictors we think are important to the outcome.

We have to collect and transform first-party and third-party data sets for this to work well. We need to be able to thread together the identity of an individual to use as our historical test cases. It comes down to having a well-thought out data strategy for collecting, storing and transforming data before any #dataanalytics has started.

Keep up the great work organizing, socializing and executing a solid #datastrategy. I’d love to hear how you are able to build capabilities that solve #analyticsmaturity for your organization.