Success with analytics projects
How often do we see clickbait about data science projects doomed to fail? Why do so many people relate to failed data projects? It’s a challenge to implement data solutions, no doubt. Having led many projects over the years, I keep a process in mind to increase my likelihood of success. Each step can trip you up in its own way, as can skipping them. If you relate to the doomed analytics narrative, you should consider the involvement of Operations, the presence of a good Data leader, and Product Management or Sales skills you may be missing.
STEP 1 - Identify a business outcome to improve
When I was an individual contributor and newly hired in a former role, a data scientist on the team asked me what I wanted to work on. This is not the way to start a project. Leadership needs to have insight into top priorities and direct resources to work on projects tied to specific objectives.
STEP 2 - Determine the data solution to address the problem
One of the best ways I’ve rounded out my skills was to dive into the subject of Product Management. It’s still uncommon to have Data PMs on a team, so it’s an entirely useful subject to be familiar with. Even with PM attention, there will be multiple technical solutions to choose from. Keep in mind the concept of optimization with an eye on the business outcome to improve.
STEP 3 - Define how Operations will utilize the solution
This might be the most common point of failure in analytics projects. Data teams can’t solve problems in isolation. The solution may be built but if it’s not utilized then value isn’t realized. In parallel to designing the technical solution, engage Operations to define how processes need to evolve.
STEP 4 - Illustrate the potential improvements
How do you keep leadership in the loop and generate excitement about your work? Once a solution is designed, you’ll have a better idea of expected improvement over the baseline. Even if some assumptions are required to illustrate the concept, a short presentation at this step is helpful before the bulk of the work gets going.
STEP 5 - Implement technical and operational solutions
Building a technical solution might be the easiest part of the process. How many data professionals are introverts who’d rather put their headphones on and disappear into coding for a couple of weeks? Occasionally, challenges do arise. Effective Data leaders confront issues and inspire cross-functional partnership. Individual contributors shouldn’t be expected to manage it alone.
STEP 6 - Present, educate, measure and iterate
Some people are excited about the use of data in their non-data roles. Some people question your motives in pushing data solutions. Analytics maturity can vary within a company. A presentation of expected improvement is necessary to gain acceptance, and also serves as an important project artifact. Think like a salesperson! Be open to feedback and emphasize the desire to measure results and iterate. Your work adds value when it’s fully implemented in the business, and this step is the final mile towards success.