Interview Series —Stephen Yu

Hosted by Career in Analytics

Decision-First AI
Course Studies


Welcome to the next installment of Career in Analytics interview series. This forum is designed for decision science professionals — both beginners and veterans — to meet one of our members and engage in a conversation with them. We want our group to be a place for great conversation and debate.

CiA: Welcome, Stephen. Can you tell us a little about yourself?

Stephen: I have been in the data and analytics business for about 30 years. I’ve even met young candidates who boldly told me that he never thought that there was such a thing called “data mining” before this century, but there have been many data professionals from the past generation. We just had different titles and computers were much slower, but we have been carrying good disciplines in dealing with data and analytics.

Currently I am Practice Lead of Analytics & Insights in a premiere outsourcing company called eClerx, based in Mumbai, India. eClerx helps marketers and decision makers with a full spectrum of digital services, from content production and management, digital production, data management, and all the way to advanced analytics. My role is to manage onshore consultants in U.S. and U.K., and to align our analytical services to best match client needs.

Previously, I served as the head of analytics for Infogroup, one of the largest data compilers in the U.S., for over 7 years. Before that, I was the founding CTO of I-Behavior, which is one of the most advanced behavioral targeting company with data from over 2,000 sources under analytics-ready architecture, now part of WPP organization.

CiA: What does analytics mean to you and your company?

Stephen: For eClerx, which is a service provider, analytics means our future. Clients are much more demanding these days, and they are definitely NOT satisfied with one dimensional reports. They want to understand what all those numbers mean to their businesses, and what they are supposed to do next. One may call it simply “insights”, but deriving meanings out of mounds of unstructured and unrefined data is not an easy task, involving ample amounts of data management work. But in the end, the process of putting all those data in the business context falls under what we casually call analytics. And if we are not serious about analytics, we won’t be able to serve our clients at a higher level.

For our clients, analytics means survival. Playing the role of analytics consultant, I see great variances in adaptation of analytics in organizations. And the differences between organizations that are seriously into analytics and the ones that are not even considering using analytics in their daily decision making or target marketing are growing larger and deeper by days.

Simply, modern businesses that are for profit must stay relevant to their customers at all time. We are living in the days of information overflow where average individuals are exposed to at least 6 or 7 different types of screens every day. That means, the buyers are more distracted that ever. To cut through the noise and stand out, marketers must be really selective in terms of whom they contact through what channel, and once they decided to contact someone, they need to know what to offer. Analytics provide answers to all those questions. Ignoring that fact could be a fatal mistake for the survival of the organization. In short, analytics is not an option anymore, especially for companies that sell products and services for profit.

CiA: Can you provide our forum with an analytics challenges that you’ve come across?

Stephen: Does it have to be just 2? I think I can write a book about it. In fact, I have been running a series in Target Marketing Magazine about that subject for 3 years now.

But if I have to pick just 1, I’d say size and state of data. Simply, most of datasets are utterly inadequate for any type of advanced analytics. That is why most data scientists or analysts spend more than 80% of their valuable time fixing and realigning provided data. Sad truth is that data hygiene and summarization work requires different type of training all together, and only a fraction of so called data scientists or statisticians are really good at data manipulation.

Of course, there are movements where developers are trying to create an almighty machine that just ingest “raw” data and spit out answers for the decision makers. But I dare to say that even such machine would have to have separate modules to refine the data to be used in advanced analytics. “Garbage-in-garbage-out” is an old term, but faster machines and bigger datasets won’t overcome it that easily. Not with a brute force for sure.

That is why data players must take an holistic approach, and must not take the pre-analytics steps lightly. And if one develops a truly “analytics-ready” environment, applying the resultant algorithms and rules to the whole universe will be much easier and faster on the backend, as well.

The 2nd most common challenges in the analytics business is lack of proper resources. “Data Scientist” seems to be title du jour, but let me be just blunt here; many of self-proclaimed data scientists are posers.

This data business has been a team work, and it always will be that way. And such teams are generally made of:

  1. Business analysts/Research analysts
  2. Database developers/master data manipulators
  3. Statisticians/mathematicians

Now, it is really difficult to achieve a master’s level proficiency in just one of these areas. Successful analysts are great at 2 of these areas. Folks who are at that high level in all 3 areas? Maybe 1 on a few hundred analysts. And the media are saying that industry needs tens of thousands of data scientists, right now. No wonder everyone is having hard time filling the role of such super-duper analysts.

If we look at the data business as a team sports, things become much simpler, though it is still not easy filling in just 1 of these roles these days. And the managers who wants to hire just 1 super analyst and call it a day? Good luck finding such candidates, and even if they did, good luck motivating them continuously with that attitude.

Maybe we should just wait for some AI to take care of at least #3 role here? For more details, search for my article titled “How to be a good data scientist by Stephen Yu” on the Internet.

CiA: What are the biggest analytics mistake you’ve seen people making?

Stephen: Analysts often forget that analytical work is to make differences in the real world. One may have found some amazing correlation between 2 seemingly unrelated variables. She should be able to ask herself “So what?” question. Is that discovery just an interesting tidbit at a party, or is it a million dollar idea?

In other words, all data and analytics activities should start with the proper business goals. Cracking down terabytes data might be great fun, but we should be able to pull the plug if the work is not deemed beneficial to businesses and/or fellow human beings.

Once the goals are set, data must be re-arranged for the purpose of analytics. Too often, I see that analysts are working within the boundary of inadequate and incomplete datasets, for technical and political reasons. Worse, business people make decisions within the limitations of such analytics. The order should be completely reversed. Business goals must set the tone and the specifics of analytical activities, and analytical requirement must control the way the data are managed and realigned. Setting this order straight must be the job of the leaders, and conducting analytics within the context of the business goals is the job of the analysts. Never the other way around.

CiA: Do you have any career advice for aspiring data scientists?

Stephen: First off, one must be good at data management or statistics, if not both. Then the hardest part will be sharpening business acumen. Unfortunately, business insights do not just happen by looking at the same report for weeks at a time. And without strong business sense, it is so easy to remain as a data plumber, who just move data from here to there, but not producing relevant and actionable insights.

My advice is to get out of the world of data, numbers and mathematics, and actually live a life of a consumer. Travel the world and see how different market places are. Question everything, as the definition of “transaction” should be challenged.

Be a good consumer and remember what the differentiating factors between buy and no-buy were. Look at marketing messages and commercials from the promoter’s point of view, not from an annoyed consumer’s point of view, and try to trace back all the decision making processes that “they” must have gone through for you to see the commercial at that particular moment and that particular channel. As none of those happened by accident.

Imagine starting a company of your own. Soon you will realize that having a kick-ass idea and some coding skills are not enough to launch a profitable business.

Finally, be curious about someone else’s job at the work place. Valuable lessons are everywhere; maybe you are just not taking them in.

CiA: Thank you, Stephen. As always, we will now turn things over to our members and see what questions they have.

Career in Analytics is a forum dedicated to connecting beginning analysts with experienced and veteran mentors. Our topics cover a variety of interests in the area of analytics and professional career development.

We would also like to thank — Corsair’s Publishing for their help in bringing this content to you!



Decision-First AI
Course Studies

FKA Corsair's Publishing - Articles that engage, educate, and entertain through analogies, analytics, and … occasionally, pirates!