Today’s business executives are increasingly applying pressure to their human resources, sales, customer service, IT, and finance leaders to use dynamic, predictive analytics. It's easy to waste time on predictive projects that deliver little value. Here are five important steps to follow when selecting a predictive project.

Identify a Business Problem to Solve

Almost everyone working on modern predictive analytics discusses the need for a defined business problem before engaging in a predictive project. And yet, the number 1 thing I hear when I speak with businesses is, “I need to do a predictive project, but I don’t know what to work on.”

Without a specific problem to solve, your analyst or vendor will do nothing more than crunch data, hoping to find something interesting. Crunching data without a specific objective is a very expensive, and typically a very unproductive, use of your company’s time and money.

Consider these examples:

Imagine searching for a house on the Internet before you’ve decided what kind of house you want. You’ll find some interesting ideas, but nothing that will make you act on buying the house.

Imagine reading Wikipedia looking for some “truth,” without defining what question you’re trying to have answered. You’ll find some interesting ideas, but likely none of them will give you the truth you’re looking for.

You need to first know what you’re looking for before you embark on a predictive project.

Decide if You Want to Solve an HR Problem or a Line-of-Business Problem

Examples of HR problems include predicting who is going to retire, which training initiative will yield the highest attendance, how your current hiring processes will affect discriminatory hiring practices, or predicting your company’s requirements for engineers.

The benefit of solving an HR problem is that you’re solving a problem that is meaningful to your department (and typically only to your department). The downside is that the rest of the business will be less excited about the work you’re doing to solve an HR process problem. Also, it is harder to quantify the business impact of solving an HR process problem.

Examples of line-of-business problems include predicting which job candidates will be top sales producers for open sales roles, which call center reps will be a good fit as a call center manager, or which truck drivers will be in more accidents.

The benefit of solving a line-of-business problem is that it is as meaningful to the entire business, because it is likely to affect revenue or cost. This will get your project much more visibility and additional resources for ongoing predictive work. The downside: You are likely to get many more requests for predictive work from the business after they see the kind of work you can do that affects revenue and cost.

Combine HR Data With Line-of-Business Data

If you are looking to predict and solve a problem in the line of business (such as boosting sales, reducing errors, or increasing calls per day), the outcome data in the line of business exists in software systems in the lines of business, not in HR. For example, sales performance or calls per day data exist in the sales operations department or the call center or some other non-HR database.

You can’t predict which sales candidates are going to make their sales numbers without sales data from the sales department. You need to use line-of-business data as well as HR data. Unless you only want to predict something that affects HR, you’ll need data from the line of business as well.

Go Beyond Predicting Trends

Individual predictions deliver the greatest return on investment (ROI). Many departments have been forecasting trends for years—and, in fact, many predictive projects we hear about are old-school forecasting projects. You need to move beyond forecasting to deliver the kinds of ROI that excite your C-suite.

Forecasting examples include:

• forecasting future inflation rates

• forecasting product demand

• forecasting workforce trends over the next one to five years so you can plan

• forecasting sales next quarter or next year.

While forecasting is extremely necessary, it is quite different from modern predictive analytics initiatives. To reap the ROI of modern predictive capabilities, organizations need to move to predicting to the individual.

Examples of predicting individual outcomes include:

• predicting which specific job candidate has a high probability of being a top or bottom performer

• predicting which specific customer prospect is going to click the coupon and buy the offer

• predicting which specific supplier is going to go out of business.

The ability to predict to this level of granularity should be the goal of modern predictive projects. ROI is higher because it helps your company to take specific action with high-cost or high-revenue potential targets.

Use Predictive Analytics Early in Your Processes Before It’s Too Late

Many companies focus on predicting the flight risk of existing employees as an early project. This reminds me of a bank predicting which loans will fail after they’ve already loaned money. After the relationship is extended is the wrong time. It’s too late.

Modern predictive analytics allows you to predict before you commit. That’s the point. Predict before you make the mistake. Banks put a lot of effort into creating predictive models that predict your probability of paying or defaulting on a loan before they extend the loan.

With these five steps, you can start to maximize your own predictive analytics project in alignment with your entire organization’s strategy.