February 18, 2016

How Patrick Kennedy used Dataquest to transition into data science

To highlight how Dataquest has changed people’s lives, we’ve started a new blog series called User Stories where we interview our users to learn more about their personal journey and how we’ve helped them get where they needed to.

In this post, we interview Patrick Kennedy, Consultant, Author, and Entrepreneur. Patrick has held executive roles at a variety of real estate firms. He’s now learning data science because he wants to explore its impact on existing industries. He has quite a few ideas for data science companies and projects that he’s working on.


How has Dataquest helped with your learning process?

Hard to put into words. It has been phenomenal. It is exactly what I needed to bridge the gap between my former life as an experimental psychologist and my current life as someone focused on building businesses. I knew much of the statistics and some of the coding but was completely out of the loop with the model building techniques housed in Sci-kit learn. I was stuck knowing that a harder look at data would yield great results but was still manually building the applications to tie our financial reporting to our sales pipeline to our individual sales person goals to our budget to our performance the previous year. It was a bear. Now it is terribly exciting to think about what I can produce with these types of models. Almost more importantly though, it has shown me a new way to think about developing intelligence.

What made you decide to start using Dataquest?

I learned about DataQuest from a recruiter at Galvanize. I learned about Galvanize from a meeting with another large commercial real estate firm that transacted their deal here in Austin. Once I heard about Galvanize’s offering, it formed the catalyst where I knew my future lay in connecting data with business in a way that I was not experiencing in my former position. The recruiter at Galvanize passed me to a quora post by Galvanize’s CTO who mentioned DataQuest as a great tool. I laid out a plan to tear through as much of the DataQuest content as I could, as fast as I could. It has been a struggle due to all the new content you all produce!

You’re doing a General Assembly data science course, and will soon do a Galvanize immersive. How would you contrast the bootcamp experience with the online learning experience?

I see it as a stepping stone. From everything I have read, the data science immersive is going to kick my butt. However, I get my feet wet with DQ so I can know what I am doing. I can explore a variety of different topics and I can do so in a structured environment. It is the best way to learn. With the introduction of the guided projects, DQ takes the training wheels off a bit but still provides the basic outline of how to organize your thoughts in a data science process. General Assembly provides a little less structure and a little more responsibility to figure it out on your own with a capstone project that you pick, investigate and present on your own. GA provides assistance and teaching all sorts of aspects of data science in a hands on way to allow one to succeed in developing their project but ultimately it is up to the individual to get after it. Galvanize seems to be throwing the individual in the fire and seeing what comes out cooked on the other end. With adequate preparation before hand through DQ and GA, I think it is a perfect 1-2-3.

What’s your general data science learning process and plan?

My plan is two-fold. First, I am poking at a variety of different business ideas to see what may launch, each with an aspect of data science to it, and some more than others. Having sold my last business I’m lucky to have the freedom to poke. Second, and knowing that it is quite difficult to get a business afloat, I am leveraging Galvanize’s outcome assistance to identify some good spots for me. It is always nice when a group is contractually obligated to place you. In the end I don’t imagine I’ll be doing just one thing – I get bored too easily. I may even just write another book on the process of learning data science.

You’ve had a variety of career and educational experiences in psychology, real estate, and management. How have these lead you to data science?

Well it has been an interesting path so far – I wish I could say it was all premeditated. But the real feel for data came as a PhD candidate in experimental psych at Columbia. It was the first time I really dug into specific tools to analyze data as well as building software applications for experiments and structuring them in a way to make data collection easier on the back end. It forced me to think about data in a soup to nuts way. Plus being around a cohort of individuals all after the same sort of goal allowed me to figure out different ways of doing things. My good friend Dave taught me about hierarchical mixed models and Baruch taught me how to analyze small data sets in meaningful ways. The trouble for me is that once we had the insight, the goal was to publish, promote and move on to the next study. No part of that involved helping others. My research was on the feeling of being in the zone with aspects I was trying to involve in video game design, peak performance and employee engagement or burnout. Time and time again I was told that my focus should be on research not applying the research. After meeting with a (now former) advisor about all the interesting applications the in-the-zone research could touch upon, she told me she knew nothing about business and that maybe I wasn’t cut out for research. So I took her up on that and left in the midst of pulling together my dissertation proposal.

Unfortunately for me I threw the baby out with the bathwater and tried to avoid strict data analysis at all costs, thinking that it was relegated to those Ivory Towers. Instead I got into the start-up world, launched a couple of my own (one became a zombie, the other burned up quick), got a business degree and worked at a technology incubator. At the incubator I kept seeing these companies I’d help refine a pitch deck (made easy thanks to my years of teaching at Columbia) go on to raise a good amount of funds and graduate. But I wanted more than just to usher a company into an adolescent stage. I wanted to be in one of those companies. I did my best to weasel in but was told I could either be sales or engineering. I didn’t think of myself an engineer and sales put a sour taste in my mouth so I did what any non-sales, non-engineering person does: consult.

One of the companies I consulted with ended up hiring me to join the firm as a director of operations. It was a commercial real estate firm – something I had only a vague idea about. What they needed seemed simple enough: help figuring out who did what. It was confusing to me that this was even an issue but when you try to grow fast without an operations backbone you miss out on some of these structural components. So I stepped in and immediately saw an application for all that research on being in the zone I conducted years before. I could use my research on employee engagement to grow a culture and see how that may affect revenue down the line (hint: it affects it significantly).

So here I was using my old research and it was working. It prompted the question for me – what else can I do here? I worked like crazy developing all sorts of tools and procedures that resulted in my first published book, my promotion to COO and ultimately transacting the sale of the company to a global firm. It was awesome. Stressful, but awesome. And it was made by an investigation into all sorts of data. Building a real time sales funnel requires data collection, munging, analytic and presentation techniques. I was manually doing much of it but continued to find new and improved ways to automate. At about this time however things started to get more friction.

The more I tried developing scalable techniques for the ultimate goal of being able to both predict a prospect’s likelihood of signing a letter of intent with us and predict which broker would best maximize that likelihood, the more resistance I felt. While I was put in charge of three offices sales teams of about 80 people, no one really cared about maximizing revenue. It was an oddity that took me a while to wrap my head around. The short story is that brokers typically make more than 50% of a sales comlesson allowing for a 3-5 year old broker to be making half a million dollars a year. Given brokerage firms typically do not have much money left over after giving that large of an amount of money to the broker, this results in little to no management, or management by broker. And one can easily assume what will happen if a broker/manager can make millions of dollars a year brokering, what they will spend their time doing.

I was fighting a losing battle. Trying to push brokers to make more, when they are able to play golf 4 days a week and make half a mil a year. Why would they want to work instead of golf? Sure they all recognized it was important but it was important for somebody else to take part in the system. Not them. As with many ideas, my frustration led me to once again realize the power of data. But data needs a context in which to thrive. The data I was generating, manipulating and presenting was having little to no effect. I needed a switch.

There is considerable data involved in any sales organization from tonality, sentiment and word choice in an initial cold call to predictability in pipeline management to recommendations for site and office space location. Heck there is abundant room in investigating negotiation techniques. I decided that instead of fighting all around me for these more old school brokers to turn a new leaf, I wanted to learn whatever I could as fast as I could to investigate all these potential data sources. I wanted to know how to collect, how to manipulate, how to get that data to talk to me in a meaningful way.

This is what brought me into the data science fold. And yes, this is a super long answer to the question.

You have a ton of experience with business operations and thinking about businesses at a high level. How could data science help you with that?

Data science can operate at almost any level of a business. It could even run businesses, but that may be just me being influenced by this book I just finished reading (Darknet by Matthew Mather). For operations, specifically, it is huge. Whether you are running a business with heavy supply chain management or a predominantly sales outfit like I was in, data science is a crucial part. Partly because the term data science encompasses so much but also because there is data everywhere. Each successful application of a data science technique ought to result in at least two more opportunities for application. As a for instance, we had used a CRM to manage all of our data. But after spending more money than I care to admit, we had to hire someone to manage the database because no one wanted to use it directly and 99% of the time it was used merely as a digital Rolodex. RelateIQ had done a great job of using natural language processing techniques to collect data from a variety of different sources (calendar, email, VoIP phones, etc) so that the data entry aspect of a CRM was obviated. There was another tool (whose name I now forget), served to give you the likelihood of making a deal based on your email chain. Now LinkedIn has their Sales Navigator tool which makes their entire platform a CRM.

With internal operations, sales is a no brainer but management has plenty of opportunities for data science applications. Goal setting is huge in organizations now. Rather than standard performance reviews (which really only serve as a paper trail when you are angling to terminate someone), goal setting allows a manager to build rapport with their employees and encourage them along their path. Most goal setting “programs” in organizations are housed in excel files and the information lies dormant. Instead this information can be fed into applications that provide assistance if an employee is stuck on a particular project and hasn’t updated their status in a while or alternatively that recommend collaboration with others if other people are working on similar applications unbeknownst to them. Much of management is inspiring employees and facilitating conversations. Both of these roles can be automated to certain degrees.

What are the biggest misconceptions that you see out there about learning data science and getting a job?

The first is that it is a fad. There are many people here in Austin I have spoken to that roll their eyes when they hear the term ‘data science’. I get it. It is a hot new thing that companies say they need even though they don’t really know why. Sure that is annoying. And statisticians, computer scientists and engineers have known aspects of these techniques for quite some time. I was surprised when I learned much of the data science model building is done with regressions, considering that regressions were the models I used for most of my experiments! I just never knew the relation between model building in a experimental context and model building in a business or engineering context. But underlying this is that data science itself is not a fad. It is the reorganization of a variety of different disciplines. That is progress! To be able to take somewhat unrelated fields and put them together is great. Sure the fervor may calm down a bit in the next few years but the importance of the work done in the field won’t.

The second is that it is difficult to learn. Some people say they don’t get code. Others say they are not math people. And sure if you are going to want to be the best data scientist there ever was, you are probably going to want to know a bit about code and math. But you don’t have to be the best. You just have to be good enough to get your goal accomplished. So for those who are genuinely interested but are scared by the new language you will have to learn or doing math problems, think about what you want to be able to do. You want to build a new type of movie recommendation service? Start reading about services out there, what a recommendation is made from and what those recommendation services are coded in. Break it down into chunks. Don’t start by picking up a book on Java or Python or whatever. You’ll get turned off pretty quick because there is no goal pulling you forward. I made a mistake when I was a guitar tutor in college. I was tutoring a boy about 9 years old. He wanted to play Elvis songs. I instead told him that he needed to learn his scales first so we spent the first three sessions of 30 minutes just going over scales. His mom called me up before session four and said, sorry, he’s not interested anymore. I feel sorry that he may have lost interest in a musical instrument because of me but at least it taught me a valuable lesson: let the goal drive performance.

What are you most excited about learning in 2016?

To be perfectly frank, I want to learn how to win a Kaggle competition. I’m in the midst of one right now and am throwing my limited knowledge at how to build a better and better model. But phew, there are some good people doing these competitions! I have a hard time turning my competitiveness off… I want to win 🙂

Beyond that though, the most interesting aspect of data science to me is natural language processing. So much of how we feel, think and interpret others depends on language. A friend of mine did a clinical psych degree at Harvard where his research showed mobile therapeutic interventions helped alleviate anxious and depressive feelings. Can you imagine Siri as your therapist? There is a lot of good to be done out there and I think we are at the cusp of something great.

Vik Paruchuri

About the author

Vik Paruchuri

Vik is the CEO and Founder of Dataquest.

Learn data skills for free

Headshot Headshot

Join 1M+ learners

Try free courses