Large companies tend to play it safe, and that’s often an understatement. When it comes to getting funds approved for a new project, most companies have a formal process that must be followed. This process typically includes detailed assessments of the costs, risks, and benefits of the proposed effort.
Large companies tend to play it safe, and that’s often an understatement. When it comes to getting funds approved for a new project, most companies have a formal process that must be followed. This process typically includes detailed assessments of the costs, risks, and benefits of the proposed effort. This approach works great for many purposes, but not when the goal is to innovate with analytics. To successfully innovate through analytics, consider the portfolio model that I’ll discuss here.
If a consumer goods manufacturer wants to add a 20th flavor to an existing product line, the costs for doing so can be very tightly estimated. After all, the company has already gone through the process of deploying a new flavor 19 times. Similarly, if an analytics team is asked to build yet another propensity model for a new product category, it is very easy to estimate the effort. In addition, the criterion by which the success of the effort will be judged (model lift) is clear cut from the start. There are few unknowns in either of these scenarios and getting through the typical corporate business justification process is easy.
When some totally new analytics are desired, however, the typical process breaks down quickly. Assume that my analytics team is asked to apply newly available sensor data from engines to develop algorithms to predict part failures before they happen. This is much harder to fit within a typical approval process. First, the data itself is not yet understood because we have not explored it to understand what it contains and what data quality issues may exist. Second, we don’t yet know what part failures we’ll be able to effectively predict and we don’t know what metrics will be needed to do it. Last, we won’t know exactly how to measure the success of our findings until we know what we’ve found. In a nutshell, there are a lot more unknowns in a scenario pursuing innovative new analytics with new data than in a scenario pursuing more of an already understood and implemented approach.
Given that there are a lot of unknowns and a lot of perceived risks, how can an organization ever get projects aimed at analytic innovation approved? The answer is to change the funding model from a project based funding model to a portfolio based model. Traditionally, each project is justified by itself. The project has to limit risk to acceptable levels and it won’t be approved unless success is, if not assured, then highly plausible. This approach is incompatible with innovation, which by its very nature requires risk.
What is a portfolio funding approach? It is simple. Instead of approving individual projects that are very low risk, a portfolio of higher risk projects is approved. Accepted from the start is the fact that many of the projects in the portfolio won’t work out well. However, the few that do work will be of high enough value to make up for the losers. This is no different from how venture capitalists make their money. Even the most successful venture capitalists lose all of the money they invest in most cases. Similarly, even the best baseball players in the world only hit the ball on 1/3 or fewer of their attempts. The idea of funding analytics as a portfolio really isn’t unprecedented or extreme.
So how does it work in practice? An organization can assign a small team to be focused on discovery analytics. Note that I am only suggesting redirecting a small portion of resources, not all resources, to the new model. Instead of being measured on each project, the team is measured on a yearly basis. To get started, a list of high potential analytics is identified. While there may be a lot of uncertainty as to which of the ideas will work, there is confidence that at least some of them will work based on the experience of those compiling the list. Which ideas will be winners will not be clear at the outset. As the team works down the list, the concern isn’t which ideas succeed or fail. What matters is that enough of the ideas succeed to pay for the team’s efforts. When smart people are given a solid portfolio of ideas for new analytics to pursue, the risk of the portfolio is far less than the risk of the individual ideas.
Pursuing innovative analytics through a portfolio funding model isn’t about removing accountability or financial discipline. It is about applying accountability and financial discipline in a way that accounts for the realities of the situation. It is also about providing leeway to the analysts tasked with discovery and innovation to truly try new approaches to improving a business through analytics. Sticking to a classic project by project funding model will ensure that an organization achieves primarily incremental improvements over time. Adding a bit of portfolio funding to the mix can help drive real innovation and change through analytics.