With all the hype around artificial intelligence, machine learning, and deep learning, companies around the world are displaying a desire for a data play. Or in some cases, an artificial intelligence play. And while the perspective is admirable, often it comes at the expense of basic groundwork. This includes, but is not limited to, the lack of ETL, the lack of instrumentation beyond operations, a plethora of unorganised data, and even worse, no infrastructure. In our experience, the most common reason for this failure from the get go is the gap between idea and execution, fuelled only by questions from article headline reading customers and investors that want a stake in a product that is advanced without truly grasping what that entails. So before building a data science team, consider the following.1. Why Beyond “my investors are asking” and “our customers want to know”, there needs to be a greater ‘why’ for investing in a data team. Your business needs to have a set of predefined and evolving goals in place, all of which are long term. In the digital age, long term – at minimum – means five years. It was with a long term approach that we helped 800Flower sustain its revenue channel through search and through which we built a B2B content marketing strategy for Insights. So be unafraid to have a sit down to figure out the state of play. A short term goal of getting investors off your back is an unwise reason, internally and for the poor scientists that will get roped into the uninformed mess. You need to also define what the scope of work will be for the data science team that is roped in. Data can be engineered, analysed, and communicated. Recognise how high this will go and plan ahead based on the tangible goals in place that justifies the existence of the data science team. In the initial period, the experienced data science subject matter expert that your company recruits, will be spending half of his or her time in identifying, qualifying and interviewing potential team mates. Let them. 2. When It would be best to hire a data scientist after an infrastructure is in plan, an ETL is in order, and when there exists instrumentation beyond operations. If your decision makers know what data sets need to be collected and why, it’s a sign that the mindset ground work is in place. This will also determine where in the organisational chart the data science team will be placed. 3. Culture Let the subject matter experts be, take their recommendations at heart, spend time working with them, and understand at least the basics of what they are doing. If you’re not committed to doing at the very least this, without being a cog in their sid, give up and wrap up your dreams of building a data science team. A truly smart company treats data and its scientists as first class citizens. 4. Infrastructure If there is none, be upfront about it, lest you prefer mass resignations on the first day. Be clear about where you are, where you would like to be, and attain estimations on when reaching that point will be. In an ideal setting, data scientists would like to dig into their role from day one. Avoid a mismatch in expectations vs. reality by being upfront on interview day one. 5. Retention Failure to do any of the above can be catastrophic in retaining the talent that has been acquired. A data science team lives for the ETL, not for setting it up or being told how to do their jobs, least of all by a clearly non subject matter expert. For the most part, your offer must be great in the work being demanded rather than the monetary compensation being offered. A great culture around the former encourages scientists to stick around, as is the acceptance of their own interpretation of work life balance. Just like with any other team, the data science team is interested in a working environment that offers unique goals, a big picture mission, and career growth. Now you can consider the above and avoid project failure and team burnout.