BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

How To Prioritize Core Business KPIs To Accelerate Value Realization From Data Science

Forbes Technology Council

Zohar Bronfman is the CEO and co-founder of Pecan.ai, a predictive analytics platform built to solve business problems.

Machine learning and data science can transform businesses with valuable insights that directly impact the bottom line. Still, many companies fear incorporating data science because it’s thought to be a complex, expensive endeavor.

Today, many companies see their needs, business goals and data as unique from every other company. I’ve worked with companies that wanted to incorporate every data set they could access—then wasted half a year validating all their data.

The truth is, your organization is not that unique. You don’t need to manually review every data set or handcraft every model to incorporate AI-based predictive modeling into your business processes. For companies looking to incorporate predictive modeling, here’s how you can achieve results by starting with essentials and moving forward strategically.

Trust the people who know the data and understand the questions.

First, start with the business problem that you’re trying to solve. Focus on what matters most and prioritize use cases that directly impact revenue. Set aside the idea of creating a large data science team, ingesting all possible data and building the “ultimate” model of your customer. This often wastes time, primarily because customer behavior and market conditions change rapidly. By the time an “ideal” model can be completed, customers have adopted new preferences and habits, especially in today’s fast-changing environment.

To evolve from data insights to business-critical predictions, the business case must come first—not data science. Data science initiatives can fail when we ask data scientists what we can learn from the data to resolve business dilemmas. In our recent survey of marketing leaders, in partnership with Wakefield Research, 40% said those building their predictive models didn’t understand marketing goals, and 38% said data scientists didn’t ask the right questions about customers.

Data scientists struggle to identify useful “learning” within data if they don’t understand which questions matter. Reframe projects to focus not on what you can learn from data but on how you can use data. Identify, define and prioritize specific use cases related to your business’s key challenges.

Trust your marketing, revenue and BI teams—they’re already asking the right questions and looking for information to guide business decisions and strategies. That’s their focus. Moreover, their analysts are immersed in relevant data, unlike data scientists or consultants, who may bounce among departments and projects.

Right now, marketing, revenue and BI teams typically don’t have the skills to move beyond reactive, retrospective data analyses without enlisting external data science resources. They use BI dashboards and spreadsheets as rear-view mirrors, showing only what’s already happened. But these data professionals are perfectly positioned to shape predictive questions and proactive steps that guide the business effectively.

Focus on your core data and concerns.

For B2C companies, predicting churn well in advance, identifying future VIPs early in the customer journey or pinpointing the right offer are universal challenges, whether you’re a direct-to-consumer e-commerce company, a mobile app or a media publisher. Most consumer-facing companies use similar behavioral and transactional data with standard measurement methodologies to analyze performance and ask similar questions.

Which customers might churn in the next X days? Who will likely become a high-value customer? Which customers will likely upgrade or buy additional products with a personalized offer?

Not only are many businesses’ questions similar, but many use the same kinds of similarly organized data to address comparable challenges. Transactional data, in-app event data, web analytics or even social media data—these data sets have common attributes across businesses.

These similarities mean we can systematize and automate most data preparation and feature engineering, creating hundreds of features and then letting the model determine which are relevant for generating accurate predictions.

To concentrate your efforts on what’s essential to your company, start with the core metrics you use to track performance and adapt them for specific goals.

Marketing teams share similar challenges and predictive needs.

If you’re in marketing, you’re likely focused on acquisition, engagement, monetization or retention. See how your team or department measures ROI and the specific customer journey segment you’re focused on.

For example, if you’re tasked with acquiring customers through ad campaigns, you’re likely measured on ad campaigns’ ability to bring customers who generate X dollars within a given period. Use your data to predict the future value of the customers each campaign will bring. Then you can proactively decide which campaigns to double down on, which campaigns to kill and which campaigns to optimize further.

Modeling a customer’s predicted lifetime value (pLTV) is similar whether you’re in the razor blades subscriptions business, mobile games or streaming media. You want to understand the value each customer generates within a few days after acquisition, and then at certain intervals after that—seven days, 14 days, 30 days, 180 days and so on—depending on your business and the purchase frequency relevant to your business model. Modeling the CAC payback period for each customer is also similar for most companies.

Move forward with a future-focused approach to business challenges.

As you initiate the adoption of predictive analytics for your team, evaluate which KPIs are most critical to address. Identify which teams are most familiar with relevant data and metrics. Next, consider whether the questions that would drive your business toward that goal are similar to those asked by other companies. If so, hand-crafting predictive models may not be necessary. Instead, an automated approach based on businesses’ commonalities could efficiently address your needs. Then, begin with the core data that can drive predictive insights, rather than incorporating everything available into a comprehensive (and overwhelming) project.

Getting started with predictive modeling can sound daunting. But it helps to realize that when it comes to core business KPIs, many organizations are similar across the board. With this in mind, being less than unique suddenly sounds far more appealing. Sharing widespread business challenges could rapidly advance your predictive analytics plans, empowering you to use your data to achieve significant goals.


Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?


Follow me on LinkedInCheck out my website