Article Link: https://sloanreview.mit.edu/article/leading-with-decision-driven-data-analytics

• Companies have more data than ever, but many fail to produce actionable insights or good results
• Being "data-driven" doesn't stop analysts and leadership from making poor decisions

• Failing to ask and answer the right/meaningful questions can be a trap to fall into, data can be used to reinforce existing beliefs

• Decision-driven data analytics: instead of finding a purpose for existing data, find data for existing purpose

Asking the Right Question

• E.g. company wants to reduce churn, develops a sophisticated model to predict which customers are likely to churn, and then sends them a gift basket
• A churn model is valuable information, but the decision made off of it (send a gift basket) is untested

• How does sending a gift basket affect a customer's likelihood to churn? This can't be answered by existing data (need for experiment)

• Too much emphasis on relying on pre-existing data can lead one into the trap of only thinking of questions or context that can be answered/explained by it

• If data-driven decision making anchors on just available data, it can lead decision makers to focus on the wrong questions
• Decision-driven data analytics starts with a proper definition of the decision that needs to be made, and then looks for data to make an educated action on it

Avoid the Trap of Reinforcing Existing Beliefs

• Twitter offers a three-step attribution process for businesses to evaluate the effectiveness of advertising on it
• Get customer data > look for customers in twitter history > if customers are found, add information on their engagement with the business

• Analysts will then compare these with customers who didn’t engage on Twitter, see a stark difference (customers on Twitter are much more engaged), and conclude that Twitter plays a major impact on sales

• Twitter would love for people to make this conclusion

• Comparing customers who do and don’t use Twitter is apples to oranges
• Twitter engagement doesn’t necessarily cause sales, sales are from customers who like the product, and therefore engage on twitter
• Again, correlation $\neq$ causation

Decision-Driven Data Analytics

3 steps:

1 Define the decision that needs to be made and consider the options available

• What needs to be done and why? Is the solution already known/obvious?
• Consider all possible options for achieving the desired result, and narrow the list down through judgement/data

2 Determine what data would be needed to narrow down the list

• Develop criteria to rank the options
• Goal of data analytics is to turn unknowns into knowns
• Starting with the decision to be made forces us to consider what we don’t know
• Focus on the high-impact data that would determine the outcome of a decision
• Once this is thought out, gather the data (i.e. experiment)

3 Use the insights gained to execute on the best course of action

• It all comes down to asking important questions and thinking beyond the scope of what you already know