Advanced Analytics is not easy. It takes significantly more effort than hiring a bunch of Data Scientists and leasing Analytics software and tools.
Some questions to consider as an operating executive when preparing to embark on an Analytics journey:
- What is the mission? What do you hope to achieve? It is critical that operating executives define the operating mission upfront before investing resources.
- Is this a one-off exercise?
- Is the plan to develop a process and deploy models into production?
- Is the exercise designed to predict customer churn? Retail sales for a specific location during a specific time period? Credit card fraud detection?
- Define the mission as explicitly as possible.
- What data is available to you? This includes your organization’s enterprise data as well as third-party and public data sets.
- Which data sets consist of structured data? Unstructured data?
- Does the data reside within stovepipe applications, is it available via APIs or both?
- What resources can you leverage?
- How many skilled people with relevant experience (Data Scientists and the like) are you able to deploy?
- Will adjacent technologies be incorporated into the process such as computer vision or visualization tools?
- Consider skilled processionals required not just to build models, but to extract, cleanse and label data. Data labeling can be an incredibly labor intensive process, especially for AI and Machine Learning projects. In addition, models must be tested thoroughly before deployment into production.
- Identify the resource gap: available resources vs. resources necessary to complete the mission.
- What is the plan for when production models fail? Approximately 20% of models make it into production and eventually those production models will fail. Look no further than insurance carriers during the COVID pandemic, or quant hedge funds during the 2008 financial crisis.