Today your software solution most likely delivers the same value to your customer whether it is the first, 101st or 1,000,001st transaction. But software made smarter with predictive analytics delivers greater value to your customer with each transaction because it learns and adapts. Predictive analytics make your software solution “sticky” because your solution grows in value to your customer over time. The more the customer uses the product, the better your solution will be.
You’re probably used to hearing about web analytics rather than predictive analytics, but predictive analytics have been around for a long time. Specialty applications in banking, credit and marketing have applied predictive analytics to manage risk and analyze opportunity. Historically few software vendors outside these markets have taken up this powerful approach. But this is changing fast. Thought leaders like Tom Davenport in “Competing on Analytics” and James Taylor in “Smart (Enough) Systems” have made the case that predictive analytics are a must for operational systems.
Platform vendors like IBM (with its acquisition of SPSS), SAS, Teradata and others get it. And predictive analytics have become an essential feature in high-end CRM and Marketing applications. Recent surveys show analytics, especially predictive analytics, top of mind for CIO’s. As Forrester Research said, “IBM’s acquisition of SPSS marks an industry tipping point. In the advanced analytics segment, the deal is having the same impact that IBM’s Cognos buy had on the BI market.”
The good news for startups and small software companies is that building in predictive analytics is easy. It is often said that predictive analytics turn uncertainty about the future into a usable probability. So what could you do with this knowledge to add value to your customer? The key is to build your solution with predictive analytics in and in mind.
Here are 3 steps to customer centric software that build predictive analytics in and in mind:
- Business Step: Identify where in your solution knowing the probability of something that hasn’t happened would improve the solution.
- Analytics Step: Look at those circumstances (from test data, trials and judgment) and define what information the solution will need to calculate these probabilities.
- Data Step: Make sure the solution will capture these data with the right level of granularity so that the system can calculate the probabilities.
Let’s walk through these steps with a fictional software startup, primarily cloud based but also offering an on-premise solution. They’ve landed their first customers, and are starting to standardize and package their offering. Now it’s time to scale their product and sell to many customers, not just a few early adopters. Like most companies they have competitors, competitors who are taking advantage of the same technologies to rapidly iterate their product. The team at our startup wants their product to have a competitive edge and they know that integrating predictive analytics will increase the value of their solution to their customers over time.
Step 1: Our startup taps the expertise of a business analytics consultant – someone who can look at their solution and find the places where predictive analytics will make a difference to their customers. This important business view of analytics will tell them what kinds of predictive analytics make sense, how best to deploy them, and how this will improve the solution from a customer’s perspective.
Step 2: Our startup now knows what kind of analytics will add value. They contract with an experienced predictive analytics specialist, someone who has experience building the specific kinds of analytics they need. These analytics might predict a customer’s propensity to buy, the likelihood of fraud in a transaction, the riskiness of a customer, customer churn or the best cross sell for a customer. The predictive analytics specialist will define what kinds of data they will need to capture, and at what level of detail. This will define the right data structure and information architecture needed, how much historical data, etc.
Step 3: Our startup hires an analytic modeler (aka statistician, econometrician, or data miner). As their software solution collects data, the modeler starts to build the predictive analytic models. The resulting models, or algorithms, will make predictions when called by the software. For their local version these are deployed as code, for their hosted version they use a third party option to deploy XML-based definitions (using Predictive Model Markup Language or PMML) to the cloud. There are lots of options here, but these are the ones our startup decides will work for them.
Our startup now has a competitive edge that gets better over time. Their software solution uses predictions to give their customers a better experience, at higher value.
Your software solution made smarter with predictive analytics means greater and increasing value to your customers and a competitive edge for you.
