Why Data Strategy will be the Next Big Thing

In December of 2017, I wrote an article which touched on a few of the reasons why data science is failing. In that article I did briefly mention that data strategy is really one of the answers to solving the high failure rate of data science.  However, I didn’t mention what is data strategy in detail in that article.  This is that follow up to the article, what is data strategy and what does a data strategist do and why is it going to be the coolest job in data?

History Lesson

I have been in the data space long before it was cool. For a long time, we had your basic BI teams doing the same kind of analytics we had been doing for what felt like decades. At least in the world of web analytics, most of what we focused on hadn’t changed much since the dotcom bubble had burst. When big data came about, I saw the rise of the engineer and how many companies believed that they were competitive once they hired an engineer and had Hadoop up and running. Big problem with that was, having data and a place to store it, didn’t solve the problem of what to do with the data. In came the data scientist.

The rise of data science was a predictable next step in the evolutionary process of getting better at using data. Just having data wasn’t enough, companies needed to create insights and do analysis on their data. It was the right next step for many companies and those that were dedicated to making their data a productive asset, jumped to this stage quite quickly, which gets us to where we are today. For many of those companies, they are still stuck at the data science phase and it is not working. Essentially, they stopped evolving.

There is a simple formula that we often used a few years ago: Data, Insights, Action. The engineers brought us the data, the data scientists brought us the insights and then they tried to also bring us the action and failed pretty good at that. The action piece is not a natural fit for data science, that is a fit for the data strategist. A data scientist is very much a tactical role focused on the insights to be found in data. They obsess about tools and languages to use and focus on the algorithm that best fits the needs of the problem. But where they often fail is, what to do with those algorithms and tools, without someone to really guide the way, they often fail. It’s not their fault, they were often never trained at this part. 

I have seen data science teams spend 2-4 years and spend millions on a problem without any timeline to go into production. Or I have seen them go into production only to make wild promises of business success that never come true. Many companies think that data science is the answer they need to the problem they have. That is partially true, but only when combined with great engineering and strategy.

What is Data Strategy

Data Strategy is really the action piece of the data utilization model (data, insights, action). This role often is the missing piece to any data practices success. What does data strategy offer? In the most basic form, it is the business aspect of data product creation and management (which is what the engineering and data science pieces are building). Where as the engineer and the data scientists often need to be specialists in their area, the strategist is more of the jack of all trades.

A successful data strategist should know the engineering aspect well enough to architect systems. I don’t expect them to go out and build a Spark engine and maintain a cluster, but they should know how the pieces fit together and make very solid business decisions around what you need a Spark engine. Just because that’s the tool of choice for many is not good enough. What’s the long-term value to the company? Can you find the talent to maintain and use to tool? Is there an alternative that can do the same thing with a better ROI both from a financial and productive perspective? 

I know of a company that had spent millions on a Hadoop cluster build out. It was tons of data sets being piped in from API’s, and was quite impressive. Only problem, it just sat around collecting dust. It was an impressive cluster from an engineering perspective but they never had anyone tell them what to do with the cluster, so it just sat there as a money waster and was eventually shut down when the company hit hard times.

Next, a data strategist isn’t the person building the algorithms, but they understand the impact of what algorithms to build and why. I recently was having a conversation with a company that wanted to use deep learning to build a recommender. Yes, you can do that, however, they didn’t have anyone on staff who understood deep learning and they didn’t want to go outside their organization to get that talent. Meanwhile, they had executives demanding a recommender. This company did have people who knew machine learning and could build an algorithm fairly quickly and get it in production. During that time period they can learn deep learning while the machine learning algorithm is making money. 

This was a revelation to them, they didn’t think like this, they thought about the six months it would take to learn deep learning and how can they tell the execs to wait and to give them more money to buy all the tools and equipment they believed they needed for deep learning. There was no need to wait because the executives didn’t care about machine learning or deep learning, they just wanted a recommender because they were convinced it would increase sales. Without framing the situation in the right business context, the team and executives would have just been frustrated over lack of real progress.

More than Just Data

Data strategy also brings into play aspects that often get left out. A good strategist is also very familiar with UX and aspects of design, the psychology of the customer journey experience as well as the ever-changing areas of laws impacting the use of data. In essence, being the CEO of the data products space is the role of the data strategist. 

Many companies think they can get by without a data strategist and in some cases that is true. A small company probably doesn’t need one full time, but they also don’t need an engineer or data scientist full time. However, any company that wants to really make their data and the insights generated from the data have the best impact, you need a data strategist. 

A strategist isn’t a replacement for an engineer or a data scientist, a data strategist is the person that allows the data and the team to be put to use in the best possible way to produce the best outcomes. There are some reason great data strategists make $1000/hr.  Without this role, projects fall into the 90% failure column for data projects.  Think of data strategy as the logical next step in the evolution of data practices because that’s what it is.

Where to Find a Data Strategist

Just like before with the engineers and data scientists, data strategists are well experienced people. There isn’t a single university program teaching data strategy. Often the skills needed are learned over several years of work experience and experience in a number of roles. A great data strategist knows how to focus on driving value and doesn’t get stuck in IT limbo or data paralysis.

Your best bet is to find someone already doing this work and allow them to drain up other data strategists. Keep in mind, this job isn’t easy and just finding an MBA isn’t good enough. Someone with real business experience as a product manager or P&L owner, added with a true understanding of how data and data products behave differently than other assets. What I mean by this is that data is an asset but not like your other assets like computers, physical products or manufacturing tools. The economics of data are quite different and an intuitive sense of these differences is needed to run the practice well. This is learned over time, not taught in schools. 

I have worked with a number of engineers and data scientists who believe the data strategy work is simple and easy to do. When you are doing it right, it should look easy. I have let them do my work and quickly, sometimes within as little as two hours, they come back and tell me I can have my job back. They often walk away with an appreciation of what I do for the team.  

For me a typical day of data strategy work can involve many things:

- Working with engineering on performance improvements to our data flows and paring those improvements up with current business objectives.

- Doing code reviews with data science teams to ensure we are covering all known aspects of the business case.

- Legal reviews, keeping up on the latest changes on PII or other aspects of how data can be used. Doing work sessions with legal and security to ensure our current policies keep us in line with current data usage laws.

- Working with UX on understanding our customer journey and what data needs to be present at each touch point and ensuring engineering and data science have a good handle on this.

- Roadmap building and business case management. Often most roadmaps are technical, I do very much business roadmaps focusing on economic value to the company and relevance to the customer. Focus on ROI, conversion rates, CLV, etc.

- Data effectiveness assessments, data has a life, depending on the business case that life can be long or short.  Data effectiveness management is vital to ensuring fast performance, especially in a real-time situation.

- Data Governance, an often over looked aspect of data, but the governance of data usage is going to often fall on this person.

- Team dynamics, probably the area I spend the most amount of time. Not just managing the data scientists, developers, engineers and architects, but also senior leadership as well. Don’t assume leader knows the obvious, it’s best to assume they don’t. 

- Proof of Life testing, this is the experiments management aspect. A good data strategist is always looking at the life of tools. Engineers and data scientists will always want something cool. However what sounds cool doesn’t always make great economic sense. A data strategist should be in the thick of it, testing out ideas and tools and the viability of those ideas and tools to be in a production environment.

- Neuroscience, this may or may not apply to you but for me focused on mainly consumer or sales business cases, it does. Data isn’t just about math, data is created by a human activity and understanding how to influence that human activity is becoming more and more of a topic in many companies. Having a solid grasp of this is important.

This is by no means an exhaustive list, but on a typical day, I will touch most of these areas at least once. Data strategy is a new field, much like where data science was in 2012, it is still being defined. There are very few books on the subject and most are written from consultants from that perspective. I have been thinking of writing one more from a practitioner aspect, what do you think, is such a book needed?