Decisions made based solely on intuition, gut feelings, and years of experience, while valuable, tend to be less effective than scientific methods. Analytics provides a methodology that incorporates crucial variables and provides simulations that show the impact of our choices before we implement them.
Analytics--the application of mathematics and statistical methods to organizational decision making--has been the focus of a great deal of marketing hype. But to use it successfully requires more than products. Here's a seven-step process for applying analytics.
1. Define the problem. Assumptions must be challenged and terms defined to come up with what in Six Sigma parlance is referred to as the "operational definition." We also must define objectives. What criteria will be used to select the appropriate decision? What conditional states exist relative to the problem, and how are they affected by the solution alternatives? What is the scope of the problem definition, and how do we know when we've succeeded?
2. Identify relevant factors. Relevant factors include controllable and uncontrollable variables. Examine whether these variables are related to each other directly, or there's a range of possibilities. Be on the lookout for latent variables, hidden factors that aren't observed directly but are inferred from the behavior of other variables. Finally, factor in risk and uncertainty.
3. Focus on data collection and preparation. This step can become complicated, particularly when mining vast amounts of data. Data mining algorithms often identify features including data affinity and patterns that aren't readily apparent with a cursory examination. This is also the time to establish baselines--measurements of the current "as is" process, against which you'll compare future "to be," post-implementation results.
4. Model the solution. Here, mathematical models of a process and its variables simulate the behavior of the actual process. A working model may be largely composed of strings or chains of mathematical formulas.
6. Implement the decision. This is where models and simulations migrate from a conceptual state to practice. It requires detailed planning and project management. In too many instances, implementations fail due to a lack of qualified project management skills.
7. Follow up. Post-implementation follow-up is the time to compare as-is baseline measures with to-be, post-implementation results. Measurement periods should cover meaningful reporting cycles, such as monthly, quarterly, or annual assessment, and you'll need sufficient granularity to make needed adjustments to your models over time.
Gary Smith is a principal at Doyen, a consulting firm that specializes in project management and process analytics.