Many articles and surveys recently have identified the critical role that analytics translators, analytic interpreters or analytic storytellers play. These mediators help analytics teams, machine learning developers, AI engineers and other data scientists bring their results to life. By showing how the model works, showing how model outcomes align with the desired business outcomes, they build trust, and improve the odds of success.
The skills for this role are often described in terms of storytelling, visualization and data presentation. These are critical skills, for sure. I would add one new one – decision modeling.
Decision modeling is an analysis technique that allows you to capture the structure of a decision-making approach. For any repeatable decision – a decision made more than once following a defined approach – a decision model can be defined. These models, based on the Decision Model and Notation (DMN) standard, deconstruct a business decision into smaller more manageable sub-decisions (and sub-sub-decisions etc.). The data each of these smaller decisions need is identified and the knowledge required to make the decision – the policy, regulation, best practice or analytic – is documented. The result is a network of well-understood decision-making components that, executed consistently, make the business decision.
An analytic translator or interpreter typically uses a decision model in two ways – to frame the analytic problem the team is going to try to solve and to put the analytic solution into a business context.
Decision models are best built by working directly with business experts and SMEs. This allows the analytic storyteller to document how the business thinks the decision is made today, and how they would like to use analytics to make it better. Any storytelling and visualization is now grounded in real-world business decision making.
Decision models are also powerful tools for framing analytic efforts. With a decision model in hand, the analytics team can ask the business owners what kind of analytics would improve the decision? How might such analytics be used? How accurate would they have to be and where would they need to be executed? All this makes a successful analytic outcome more likely.
Once an analytic model has been developed, the decision model shows how it can be used. Perhaps it can be plugged simply into the existing decision-making approach. If so, the decision model will make this clear. If the business team must change their approach to take advantage of the analytic, that too will be clear along with the degree and nature of the change. Adding this understanding to visualizations will make the whole story more compelling.
If you want to teach your analytic storytellers how to build decision models to improve their stories, contact us.