Here is a thought experiment:
In 202x, AKA the year of Popeye’s Chicken Sandwich (Y.P.C.S.), Artificial Intelligence (AI) developed by researchers has been largely deployed in the business world. Operational tasks have been completely taken over by AI in particular in the field of data, Big Data, Medium Data, as well as Small Data with a capital S. The speed of gathering and processing data is no longer a challenge - decision makers can get whatever information (among public database and within their own organizations of course) within milliseconds and all they have to do is to ask the questions.
What is the total download of the App in the past 12 months? What are the characteristics of users that have a high chance of churning after the first 3 months? Among users who purchased a certain item what search terms they looked for exactly 7 days prior to the transaction and what is the correlation between the length of the term and their likelihood to revisit the site on a Wednesday afternoon? All these questions will be answered in a comprehensive way even the stakeholders have no knowledge of coding or statistics.
What a perfect world, and it reads so much cooler if you turn on the dark mode on your laptop or phone like I do. However perfect it might sound to the decision makers, it might not be great news to fellow data detectives - after all it is the truth that we seek and if the truth gets returned faster than google results, the value of analysts will be questioned.
Under this extreme thought experiment, we get a chance to examine the true value of an analyst role, and the examination feeds to a larger question: what is the competitive edge of a decision maker when the gap between information is completely eliminated?
As it turned out, having access to answers to all the questions does not create a better world. As long as it is still human that make decisions, decision makers “suffer” from psychological effects like decision paralysis which are by-products of the evolution. From an evolutionary psychology perspective, psychological mechanisms that collectively worked the best in the past got to be passed along down the way even if the individual parts are no longer serving the modern world. Hunter-gatherers didn’t need to make hundreds of decisions per day (which is also why we inherited heuristics and stereotypes) and certainly didn’t have access to all the information in the world. Simpler life back then optimized the brain to be more sensitive to emotional cues as well as high calorie food. It was not optimized for exploded information.
Having access to all the answers lures decision makers to ask all the questions they could think of. Analysts, in a way, encourage this exploration. Driven by curiosity, data analysts are prone to get to the bottom of a question. Even in real life the access to information is limited, creative analysts are able to estimate the amount of gas stations or piano orders in a city using available data points as proxies. Imagine the powerful curiosity when the limitation is lifted.
In The Sweet Spot, Prof. Paul Bloom quoted a story from The Hitchhiker’s Guide to the Galaxy where the scientists asked about “the answer to the Ultimate Question of Life, the universe and Everything” and the computer returned 42, an unsatisfying answer. Bloom thinks the reason that the answer is disappointing is because the question on the meaning of life is not a good question. “… life isn’t the sort of thing that has a meaning.” Although in data analytics, we are rarely asked about the philosophical indication of users’ behaviour, there is certainly a chance that not all questions have an answer. Knowing what questions to ask is the key because not only the additional information distracts decision making, the asking itself comes with an opportunity cost.
Back to the thought experiment, in the perfectly informed world, the value of an analyst comes from the ability to help stakeholders understand what questions they are really asking. In order to do this, they will need to plan out how they would take actions based on the additional information. What is the total download of the App in the past 12 months? If the download decreased in 2021 comparing to 2020, it might suggest a decrease in popularity and a need to update the product strategy. The real question might be concerning the relative popularity metrics. What are the characteristics of users that have a high chance of churning after the first 3 months? If they showed high engagement at the beginning but the interest suddenly decreased, then it might need some investigation into user retention rather than onboarding. Among users who purchased a certain item what search terms they looked for exactly 7 days prior to the transaction and what is the correlation between the length of the term and their likelihood to revisit the site on a Wednesday afternoon? Unless there is a specific plan for those users, this level of granularity might just add noise to the bigger picture.
But what about the small dots that might connect into inspirations? Maybe a better question can be formed only when the data is seen. First of all, throwing out questions and hope some of them will make sense eventually has a low ROI. Second of all, exploration could be exactly the goal but there still should be ideally a plan for the information.
Just ask questions. Data detectives will catch you.
