John Thume wrapped up the day with a session on Aster Operationalizing Analytics and the new Aster AppCenter. Success in Big Data Analytics requires several things he said:
- It’s all about People and Process not technology he says. The technology has evolved hugely but the people and the process they have to change remain challenging.
- Aster is a solution development tool for analytics that allows for more rapid, more flexible development of analytic solutions.
- Best practices of data warehousing remain true even as you expand to Big Data
- Governance, quality, data management all remain important
- Avoid the science experiment
He pivoted at this point to talk about the flexibility of Aster, in particular its ability to develop analytics with very little code. Less code means more flexibility, more rapid change and adaption. And Aster makes it easy he says to operationalize these things, pushing them into the front line systems that people use..
He suggests a big analytics process:
- Define business problem
- Integrate analytics into a business process
This is exactly the process I have been using for a decade – decision discovery (find and model the decision you want to improve), decision services (to integrate analytics and business rules into production) and decision analysis (monitor and continuously improve).
He correctly points out that this requires a cross-functional team that brings together the business, IT and analytic teams – the three legged stool. He also points out that you need to know what the KPIs that will measure success and map those to the decisions you are trying to improve with analytics.
In other words exactly the approach I would use for decision management and decision modeling: Identify the KPIs that measure success, find the decisions that will make a difference to your business processes, build the analytics that will improve those decisions, gather and analyze the data you need to build these analytics. Use such a model to determine who has to change to use these etc.
The Aster AppCenter is a way to package up complex analytics to make them easier to consume and use. Each app has a form that allows someone with little or no analytic skill fill out parameters and drive a potentially sophisticated analytic based on those parameters with no SQL MR. The visualization of the analytic gives you the explanation and justification you need to get people to believe it. But he correctly points out that the real deal is the UI for the end user that let’s them consume the analytic.
And that’s a wrap.