This is the second part of my interview with John Kelly where he explains the most common challenges in terms of organization and why big data investment has not yet impacted companies at scale. In part one, he answered questions relating to data science team members, how they relate to others in an organization, and what major frustrations they face when using data science tools.
John P. Kelly is the Managing Director of Berkeley Research Group, a predictive analytics practice that leverages econometrics and data science to help drive actionable and data-driven growth strategies and products. BRG empowers its clientele by applying data science to key strategy decisions being made in marketing, sales, and operations. Some examples include dynamic pricing optimization, loyalty program design, site location analysis, predicting consumer behavior, and reducing churn.
CM: What are the most common challenges in terms of organization? Do you have any advice on how to overcome them?
JK: The most common challenges depend on how far companies are on the analytics journey. There are 3 types of companies:
- Companies who were digitally native from the start and have hundreds of analysts with various skills depending on their backgrounds, from business intelligence and business analytics, to predictive data science.
- Companies with no data culture: you can find a lot of brick and mortar firms in this category.
- Companies in the middle with a small stack, and a few data scientists and database administrators who know how to use it.
Overall, I would say the first thing is culture. You have to believe you can have more by absorbing the data. That culture is defined by the sum of a number of individual behaviors:
- Cooperation: If everyone is against the data, it is not going to work.
- Leadership: There has to be someone with the courage to change the company’s mind. This leader has to know something about data science or, at least, have prior success in a data science project. Trusting in who will follow this path is also essential.
Before going through large investments, tools, and resources, companies have to start at the depth of data science. Is the data valid? Can you find true evidence of causality between the datasets? You really need a true data scientist in your company to dig the data and find relevant insights (e.g., knowing if dynamic pricing will yield more profitable results or if 100 retail locations are ideally located and the remaining ones are not).
In a top-down approach, executives would begin with the desired destination, and say, for example, “Hey, let’s look at relocating our sites.” They might also make big investments in the data in the hope that it will tell the desired story. Whereas, with a bottom-up approach, you would let your data scientists sift through the data first, and inform management of the five most impactful data relationships to the business, regardless of where they apply. Management and the data scientists could then collaborate to determine the best opportunities to pursue first, without requiring significant investment in tools and infrastructure. For this reason, I would encourage a bottom-up approach.
It takes time to support the idea of leading with the data and the data scientist... and to let them suggest what to follow. That requires trust, a data-centric culture, and a leader who is knowledgeable on how a company could benefit from data science or one with prior field experience.
I also see another challenge. A myriad of companies are selling technologies which seem ideal, but do we need them all? Even if these solutions are solving my problem right now, they might become obsolete in a few years or even months. You need to buy something that can bring value over the long term.
CM: Some companies have invested in data science, but the expected results have not come to fruition. In your opinion, what factors are to blame for this situation?
JK: McKinsey wrote a good article about this, which is their second most read business technology piece in 2015. This article states that, for most companies, data analytics success has been limited to a few tests or to narrow slices of the business. Very few have achieved what we would call “big impact through big data,” at scale.
To capture the potential of data analytics, you need a culture shift to leverage the use of data. People tend to have a lot of experience and base their insight on it. Jobs need to be re-designed to achieve large-scale benefits. It also starts with the capacity of executives to go bottom-up and not top-down and trust the data scientists to find the most impactful data relationships. If you’re in the midst of that process, you are now able to accelerate your data analysis transformation.
CM: Would you mind sharing your data science predictions for 2016? What will be the most popular data science use cases to watch?
JK: In the retail world, the use of beacon technology will provide local retail stores with massive consumer shopping data to understand the customer’s journey, both online and in-store. For example, local retailers will be able to send out targeted messages and promotions to mobile devices. I think brick and mortar businesses will start leveraging the value of data science in 2016 and natural language processing (NLP) is also going to continue to advance. As a lot of customer data is found in e-mails, online reviews, and message forum comments, it becomes more enlightening to analyze unstructured customer feedback rather than organize focus groups. There is so much we can learn from NLP in 2016. The area of cognitive thinking is also a hot data science use case for 2016 as it becomes possible for machines to save a massive amount of data, apply and observe the pattern, and then learn from history.
CM: What will be the trendy new data science job titles in 2016?
JK: Honestly, the best answer is that traditional jobs will need to start applying a data-centric approach to their job. The next generation of CMOs, Head of CRMs, COOs, Heads of Sales, etc., will need to handle data. For instance, a Head of CRM will not only have to report the current sales news but will also have to guide the force as to which prospects have the greatest likelihood to buy, and how to spot the profile of a true buyer versus someone who may be a less productive use of their selling time. The trendy new data science job titles will just be traditional job titles that are customized with some data science buzzwords.
CM: What will be the best data science tools in 2016?
JK: I am hearing noise of data science platforms that are aggregating a number of languages where you can easily add-on data, collect data through APIs, and use more push and click and drop buttons, making data science more accessible. The best data science tools will be the ones which help you the most in data preparation and allow you to re-allocate the rest of your time on data interpretation and visualization.
I will keep you posted on more data science predictions from experts in my next blog post.
On a final note, if you feel overwhelmed with sales and marketing information about data science and need assistance finding the light at the end of the data tunnel, do not hesitate to drop me a message. I’d love to help you out and get to the heart of your motivations, reservations, and desires in terms of data science!