What business questions are you trying to solve? Are you trying to predict long-term product performance or short-term risk? What data do you already have and what do you need? Not all data come from trials and tables: how can we incorporate your existing expertise and experience into models as well? What free or inexpensive data might be drawn from sources outside your business?
Creating optimal places for your data to live, typically cloud-based tables or databases that are easily or automatically updated with new information.
Data science cannot overcome poor data. Garbage in produces garbage out. For answers to be reliable, they must be derived from data that appropriately describe how you do business. This requires identification, removal, and ideally, replacement, of observations values that may be irrelevant to your needs. This needs to be done in a way that is easily described, highly-efficient, and has room to grow.
This includes the development of models that properly describe your current business outcomes and can be used to predict future outcomes, including: expected results and variability given current conditions, production risk, or the optimization of inputs. Always, the result is to produce models that are explanations that are as simple, reliable and easily explained as possible.
Model outcomes are most compelling when we can see them. Creation of graphs and maps that are immediately understood is critical to both decision-making and selling others on your recommendations.
Empowering the end-user to use their data inferences more effectively. How do they support immediate decisions? How can we develop practical tools that can be used to check future progress or test new management scenarios? This can include presentations, dashboard development, custom applications, and user training.
Ensuring that you own your data products: where they live and how they were created, and what the results mean. Making sure they are readily understood by future users and consultants.