According to Gartner, the global analytics and business intelligence software market reached $21.6 billion in 2018. The firm has also predicted that, “through 2022, only 20% of analytic insights will deliver business outcomes.” That means that organizations are investing billions of dollars in analytics with minimal return — hardly a recipe for success.

Oftentimes, this disconnect stems not from faulty data science, but from an organization’s failure to consider the activation-readiness of their approaches to real-world applications of analytics. For many organizations, activation, or the art of leveraging data to do something meaningfully different in the market, is the missing piece that bridges the divide between insight and business value.

While most mature organizations understand how to leverage analytics for knowledge discovery, far too few are able to consistently aim this discovery in the right direction. This results in undeniably impressive analytics that are functionally useless.

To avoid this trap of analytics for analytics’ sake, organizations should take the following steps when designing and evolving their analytics processes:

Prioritize High-Value Key Business Questions (KBQs) Over Pipe Dreams

In a previous HBR article, I introduced a process for arriving at the kinds of key business questions (KBQs) that set organizations up for analytics success. KBQs are forward-looking questions that establish a framework for what an organization will do with the insights produced by analytics. For instance, “Can we identify customers who churned after we discontinued one of our services, and frame our remaining services in a way that will win them back?” Or, a KBQ I encounter frequently in my line of work, “Can we map the referral relationships among healthcare providers and use our understanding of these relationships to better tailor our communications with each provider?”

Situating the KBQ-generation process in a broader discussion about activation-readiness requires taking a deeper dive into the process’ final step: prioritizing your KBQs. Once you have compiled an exhaustive list of your KBQs, you should assess them along two axes: “ability to activate” and “potential to impact the business.” (See figure below.)

Organizations that, at a minimum, understand how to leverage analytics for knowledge discovery typically end up pursuing KBQs that fall within the upper-left (pipe dreams) and upper-right (high-value KBQs) quadrants of this grid. High-value KBQs are the North Star of activation-ready analytics. Pipe dreams are questions whose answers possess immense potential to impact your business, but are difficult to act on in the market.

If you’re attempting to reduce customer churn, one of your KBQs might be, “How can we drive organic growth for our business by increasing our average customer lifetime value?” From an analytics perspective, answering this question is fairly straightforward. Armed with the right data, your analytics team can create a probabilistic scoring model that predicts the likelihood that you’re going to lose a customer early in their customer journey. However, while this model amounts to an analytics solution to a critical business question, its mere existence does not qualify the question as a high-value KBQ.

A data-driven churn prediction model is only valuable if it enables you to change what you’re doing in the market in a meaningful way — that is, if you’re able to activate on the insights the model produces. If you don’t have the right CRM and tech infrastructure in place, you’ll be unable to put your model into play in the market, and your original question will end up as a pipe dream — its potential business impact is high, but your ability to realize this potential is effectively nonexistent.