By: Soumendra Mohanty, EVP & CDAO

Sometime back, I had a chance conversation with a “Data Whisperer”; yes – a role like this exists in a large global financial institution. Obviously, I was intrigued by the role and was keen to understand, what exactly this person does for a living! This post is all about the conversation with a “Data Whisperer”, for the sake of dramatization I have taken the liberty of infusing a healthy dose of storytelling as well. 🙂

First, let’s get the definitions out of the way – A whisperer is a person who is able to tame or control animals, especially by talking to them in gentle tones, e.g. a horse whisperer. A similar role is envisaged in order to tame, control and draw inferences from messy data (read as big data)!

Data Whisperer is a person, who is able to tame and control wild data, using the appropriate tools and techniques to turn it into useful information for business success.

Wonderful! The definition out of the way, I asked the question – One way or other, isn’t this role similar to a “Data Steward” and going by the current buzz, isn’t this role similar to a “Data Scientist”?

The answer was, “Data Steward” is more or less confined to provide the business context behind the data and going by the age-old DW methodology, this role was elevated to a “conflict resolution” mechanism when you have multiple departments/functional owners defining and interpreting the data in different ways. There is a still a great degree of overlap of this role with the data whisperer role. Turning to the data scientist question, the role and associated activities got straight-jacketed into a highly technology driven definition – A data scientist only sees part of the equation. Identifying and understanding patterns and using algorithms are a great start, but they aren’t the silver bullet businesses are looking for to instantly turn their data into insights.

​I was not in agreement with this commentary, but decided to play along and understand the rationale, how so?

What does a Data Scientist do?

A data scientist’s role is to analyze large sets of data by using math, statistics and algorithms to identify patterns. The data scientist often resides outside of a business unit and spends most of his time delving deep into varied data sets to find patterns and trends. While these patterns and trends are important, the series of data outputs that identify trends don’t pinpoint why the patterns are happening. If these large scale outputs are viewed as an exact science, companies risk making mistakes because there are always outliers, and external factors, by its very definition, unpredictable.

Ok, this is beginning to make sense. In fact, in one of the big data conferences, I had heard a similar statement – “Big Data and Machine Learning algorithms is all about finding correlations. There is still a play of the business SMEs to associate causation, without which the patterns are meaningless.”

This is where the data whisperer comes into action. While data scientists can provide information and predictions based on data that they had collected, only the data whisperer can fill in the gaps to provide insights into individual thought processes and offer real world factors that throws light into the “why?” When the data scientist and the data whisperer work together, science and art combine to provide the complete view of the behavior.

The data whisperer is a human-focused analyst!

He has a fundamental understanding of the business, market drivers and customers. While the data whisperer may use some of the same modeling techniques as the data scientist, that person is tasked with combining these insights with reasoning to show what created those patterns. Unlike the data scientist, the data whisperer has the experience and skills to make inferences within the data set and determine the underlying reasons for a specific action. The data whisperer views information through a cognitive lens, adds market and business context to finally decipher why specific patterns are occurring.

When I asked for an example, he shared one of his recent whispering escapades with data. They have a state of the art “Data Lake” and a bright bunch of seasoned data scientists. They are very proud of their innovation culture that allows their data scientists to just play with the vast amount of data and in a weekly round-table the data scientists passionately present their findings in front of a mixed crowd of business and IT leads. A few weeks earlier, their data scientists detected an alarming trend of “customer churn” and they recommended a strategy of “cross-sell” to those customers. While business was still digesting the finding and impact to the business, the data whisperer delved deeper into the correlation aspects and contextualized the finding – the inference in an interesting way led to the conclusion that these customers in reality are actually not churning, but due to an earlier campaign by the bank, they were closing down certain portfolios and consolidating their wealth management options!

Well, if you think about it, the data scientist and data whisperer both play a critical role – The output analysis through pattern recognition and the input analysis of the “why” behind the behavior. For example, aggregate data may help predict customer churn by indicating that a certain type of customer may discontinue service or leave. This information can help devise an appropriate customer engagement strategy, but knowing that a customer is “at risk” doesn’t necessarily equip customer service representatives the front line to successfully engage with each individual. Imagine the situation, while the customer is trying to consolidate his portfolios due to an earlier campaign, the customer service representatives are hell-bent in brain washing the same customer with more cross-sell offers! The data whisperer’s touch adds the cognitive and influencing elements that can provide a glimpse of the hidden behavior drivers.

Things were beginning to become clearer but I had a different argument – From a user’s perspective, isn’t all data “wild” before it’s captured, curated, enriched and then we “tame” it so it can be effectively used? If so, aren’t those “taming” techniques a part and parcel of the data steward or data scientist’s role?

In an attempt to answer my question, he gave me an even more interesting perspective. During the course of our data analysis, we run into both “whispered” and “un-whispered” data situations all the time, and there’s no question that you see better results when companies treat their data well using robust data management and data governance principles like data cleansing, discovery, management, monitoring, profiling, etc. But, in reality, many times we encounter untamed, badly treated data. What do we do? We treat those data points as outliers, not conforming to the enterprise standards and either send notifications back to source systems to fix the data or remove them from our analysis. This is a big mistake! In contrast, we need to look at these wild or untamed data sources through a cognitive lens, probe deeper to find the why and then guide the data scientists to get on with their algorithms and tools.

Conclusion

As numerous reports have demonstrated, companies enamored with the notion of “big data” continue to collect data without any clear idea of how to do the analysis that might lead to insight. Perhaps, the hope is that by applying a whole bunch of sophisticated algorithms will generate the insight all by itself.

Unfortunately, it doesn’t usually work that way. Most business executives fall into camps on opposite ends of the spectrum. Either they see big data as a crystal ball that will tell them everything, or they see it as just a really big messy data store. The trick lies in defining a reasonable problem for big data to solve, and then figuring out questions you want to ask to the data you have.

Data is like horses; it can be untamed and unmanageable or it can be trained and useful. Taking data from the one state to the other can be routine and fit to the daily life of a data scientist. But then, sometimes there are wild and tricky data sets that require the skills of a Data Whisperer.

A data whisperer studies the data in action in its natural habitat, observing its confirmation at rest and full gallop, how it is fed, housed, and cared for, and, most importantly, what it was bred to do.

Certainly I see a role elevation of the data stewards, but not just confined to do conflict resolution. The data stewards with their business knowledge proximity should aim to become data whisperers, the role itself sounds a lot more exotic, anyway.