1. Decision Known/Unknown: Do you know which kind of decision you are supporting? For example, you may be supporting the decision to assign people to projects based on skill-match and availability. In this case we know what data we need (e.g., project-task schedule and skill needs, people’s skills, existing task-assignments, and availability) and what insights we need to generate (e.g., an optimal task-assignment recommendation). When the decision is an unknown, e.g., in the case of predicting sales for the quarter, insights such as the sales forecasts are generated from the underlying data about bookings, opportunities, analyst forecasts, etc. and made available to decision-makers to use as they see fit.
2. Insight Known/Unknown: When it’s known what insights are needed, methods such as correlation, clustering, simulation, optimization, etc. are used to generate insights from data. In other cases, how the data is used to create insights may not be entirely clear, and end users analyze data to find insights. For example, where the specific insight is not known, a self-service dashboard of service ticket history can be used by analysts to look for different patterns such as support trends by product, the effect of training on agent productivity, etc. When you know what insights people glean from the dashboard, code can be written to provide it on an automated ongoing basis to the relevant people, and it becomes a known insight.
3. Data Known/Unknown: When new data is explored, what the data contains is unknown beforehand. In some cases the exercise may need to start with adding new data sources, for instance to optimize truck scheduling logistics using location data from GPS, breakdown analysis using vibration sensors, etc. Once the data is understood and its cleansing processes are defined, systems can be built to provide that data on an ongoing basis.
When these three questions are asked for business scenarios that can benefit from analytics, the answers will fall within the four scenarios depicted below: