Seeing the future is the future, at least when it comes to marketing tech.  The popularity of predictive apps — tools that help marketers understand how to sell better — is exploding.  And for good reason.  Used correctly, predictive models are consistently effective.  They can help companies pinpoint the right decisions, making them more efficient; they can even be used to automate an entire decision-making process.

An early, classic, predictive model is the credit score — that familiar, numeric tool that helps lenders predict who can be counted on to pay back a loan, whether for a house to a small appliance.  Lately, the use of predictive models has grown exponentially; they’re now used to analyze a wide variety of data.  Fields that employ — and benefit from — predictive analytics include telecommunications, marketing, insurance, health care, retail, and many others.  And the list is ever expanding.

Thanks to an explosion in unstructured data — from text documents to videos — as well as improved analytical techniques, predictable analytics have more potential than ever.

Like all tools, though, they must be handled correctly.  Yet they frequently are not.

Looking for guidelines on how companies can improve their handling of this excellent tool, we at ARTÉMIA were particularly impressed with the cogent advice offered by James Taylor, CEO of Decision Management Solutions, an analytics and management consultancy, in the August 25 issue MIT Sloan Management Review.

Taylor summed up the four best areas for predictive analytics as “risk, opportunity, fraud and demand.”  Each area would probably require different models, he added: companies can’t just build a model once and expect it to have wide application. Furthermore, they may need a different one for every question they ask.

Taylor added that predictive analytics work best for operational decisions, such as determining which supplier to use.  They are less effective for strategic and/or tactical decisions, which tend to be more complex.

Taylor’s advice can be summed up as follows:

1. Focus on the decisions you want to improve.
2. Make analytics as broadly usable as you can.
3. Start with models you can scale.
4. Keep improving your models: use them repeatedly, make them better, and expand into new areas of the business.

(A link to the MIT article heads our list of readings below.)

Other tips that we have gleaned in our research include:

• Have a specific objective in mind before you start.
• Only spend time on projects that your data can support.
• Ensure that your data are consistent.
• Don’t wait for data to be perfect. Delays can kill projects; besides data scientists have methodologies that can make flawed and/or incomplete data work.

And, as always…

• Don’t rush the process!