July 13, 2018

Funding Data Science Initiatives – Part I

Data Science continues to be a hot topic.  Last year, I had the opportunity to work with a number of clients developing budgets for data science initiatives in 2018.  While each client had a unique process for developing, prioritizing, and approving budgets, many of them shared similar challenges when it came to data science initiatives.  I saw challenges arise both within defining individual initiatives as well as in assessing, prioritizing, and selecting among a collection of data science budget requests.

Investment committees, prioritization, and ROI

Investing in advanced analytics is still a new process for many organizations.  While the opportunities are exciting, there exist fundamental challenges in how organizations invest in data science projects.  This is a two-part blog post that summarizes the challenges I have witnessed and experienced across two critical players in the budgeting process:

  • The investment committee responsible for approving budget requests and the individual business stakeholders developing a specific initiative.
  • This first post focuses on the investment committee.

What should we invest in?

In many organizations, there is a central investment committee responsible for approving data, analytics, and data science budget requests.  Some organizations create a cross-functional investment committee with representation from IT, analytics, and various business stakeholders. Other organizations funnel requests straight to the CFO.  Regardless of the investment decision process, I see three fundamental issues that present challenges when trying to assess, prioritize and approve specific investment requests.

1. Benchmarking – Is it worth investing in a solution?

Most budget processes require some form of benchmarking and ROI projections when submitting a data science initiative.  The fundamental challenge I see is, while this is required, very few organizations dedicate any budget or resources to develop compelling benchmarks.  Rather than taking a data-driven approach, one that analyzes historical trends to identify the primary revenue opportunities and initial benchmarks, most organizations create ballpark estimates.  I often hear statements such as “Customer churn is roughly 30%” and I often find myself asking, “Is it 27% or 34%”. Starting any data science investment with a clear benchmark supported by data provides a tremendous headstart for any projects.  The challenges are establishing a methodical, funded approach to developing such benchmarks.

2. Due Diligence – Can we even develop a solution?

Data science requires a number of components in order to be successful.  This includes granular, accurate data to support model development, business processes that are capable of incorporating insights, and business systems that are extensible to include new metrics or scores.  This level of due diligence, something we call collaborative due diligence to signify the importance of collaboration across various stakeholders within the organization, is often executed after budgets have been committed and projects begin.  Tackling this phase and including the results in the budget request would significantly reduce the number of failed investments. Again, the challenge lies in taking a methodical, funded approach to due diligence.

3. Thanks, but no thanks – How to ensure activation?

The third challenge I see when organizations analyze, prioritize and invest in data science initiatives pertains to activation of insights.  Any data science project is only successful if the business stakeholders adapt their decision process to incorporate new insights. This requires a level of maturity within the business units that span business processes, systems, and data.  In many cases, I have seen data science budget requests receive funding with very little commitment from the business stakeholders responsible for activating the insights. Without this firm commitment prior to investing in a single project, significant risks lie in insights never being activated.  One solution is to create a form of a business contract, one that explains the expectations of the business stakeholders as well as the analytics group, prior to investing in a specific project. Tying this business contract to the investment process would reduce the number of failed data science projects.


Significant funds are being allocated to data science solutions.  While this is an exciting time in the data science community, if a significant number of data science initiatives fail, the hype will start to fade.  Investment committees can make small changes to their investment process in order to increase the probability of success for any data science investment.

Continue Reading: Funding Data Science Initiatives – Part II