The 2014 MITX Data & Analytics Summit is just 10 DAYS away! This week, leading up to the full-day event, we have some guest blogs written by members of our summit advisory board. Read this post by Amit Phansalkar, Chief Data Officer at MassMutual to get a sneak peak at some of the insights you will get at #MITXData on the 13th. Then make sure you reserve your spot and we will see you there next week!
Amit Phansalkar, Chief Data Officer, leads data science, data strategy and vision for MassMutual Financial. He heads the data venture for MassMutual to create a data ecosystem to help drive innovation in products and consumer engagement models. Mr. Phansalkar has more than 15 years of experience driving innovation and building products using big data and most recently served as the global head of data sciences and advanced analytics at Millward Brown Digital. Prior to that, he was a co-founder and VP analytics at Cognika, where he created predictive analytics products utilizing a combination of structured and unstructured data with applications in defense, ad-tech and healthcare industries. Amit is a data evangelist trying to solve complex challenges and representing complexity using disparate data sources.
How many golf balls can you fit in a 747? This simple, albeit ridiculous, question digs right at the heart of today's data analysis approaches and is the bane of business executives across all businesses.
Questions like this test the way we go from a series of assumptions about things we have limited information about, to coming up with an answer we can’t immediately test. This type of a problem, called a Fermi problem, is often used in engineering and sciences. It is used to scope a problem before attempting to build a complex model to derive a more precise answer. This more precise answer is often not needed and expending time and energy to calculate the answer can hurt the bottom line of the business.
As businesses - especially industries that serve the end consumer - can now collect myriads of data points about their consumers and are attempting to be more data driven, it becomes even more important to understand the questions that need to be answered, metrics that need to be measured and more importantly quantifications that need to be optimized so that the decisions become data-driven.
Businesses build data science teams as a way to approach this problem. But without the right mindset, they suffer with the myopia of truly unlocking the strength of this data and its ability to transform the way they approach, acquire, retain or innovate. Businesses are still pushing to answer the same set of questions with similar approaches; however, with an explosion of potential information to be utilized, they often make the mistake of disappearing in the black holes of data analysis.
It’s about an experimental approach
"Take a look at all of this data and tell us what it tells us about the segment." This is a classic question that was created by executives when measurements were somewhat restricted and only those data points that were important were measured. But when you collect vast amounts of data, it’s not a question of finding the needle in the haystack. It’s about knowing the structure of the haystack. A classic example that data has broadly shown, is that coke sales increase as visits to the beach increase. This does not mean that one causes the other, simply that these are co-related events. There is no reason to shun this finding. Can coke benefit by having more vending machines around beaches? Of course. Coke should use the findings in an experimental approach that can then be built on to ask the deeper question of, "Which locations will benefit the most?"
Most data science is about creating the next experiment with data. If one builds a series of hypotheses where the restriction is that the next hypothesis is built on the previous one, it leads to a process of discovery. This process can transform business models as opposed to create black holes where data science efforts yield non-actionable results.
Data Science is more art, and less science
When working with businesses it is oftentimes important to translate the business situation into a data problem and then translate the data problem back into a business solution. A business problem can be anything from, "We want to expand into new geography" to, "We need a tool to increase sales in a given month." To approach these problems, here are some keys I have learned:
The first step towards a successful data science project is being able to scope and size the problem, and defining the endpoint of the first experiment.
Nail down exactly what the client (the business) wants and the key metrics by which the insights become actionable.
Formalize any assumptions you have made; and make sure that everyone understands the risks.
Managing expectations is as important, if not more important, than managing your data.
It's ok to not have an expected result. Unexpected (and negative) results have more value in setting up the next experiment.
It is very important to be creative in solving the data problem, mostly by borrowing from other fields. To quote one of my favorites – Picasso –
"Good artists copy; great artists steal."