Hitmetrix - User behavior analytics & recording

A Basic Intro to Predictive Analytics

Marketers would love to have a crystal ball that tells them how their customers are going to behave and whether they’re going to buy. But because this is more fantasy than a reality, they have to settle for predictive analytics.

Research and advisory firm Forrester defines predictive analytics as “techniques, tools, and technologies that use data to find models…that can anticipate outcomes with a significant probability of accuracy.” Or, put simply, it’s taking what you know and using that insight to forecast future results.

According to fellow firm Gartner, these types of data mining approaches contain four key attributes:

  • A focus on prediction, versus description, classification, or clustering
  • A speedy analysis—i.e. one that’s measured in hours or days (or faster!)
  • A concentration on how the analytical insights link back to the business
  • An accessible and easy-to-use tool

Predictive analytics can be used for many things. As the Harvard Business Review (HBR) notes, it can be used to forecast demand for products, anticipate when technology is likely to break down, and adjust pricing and offers. It can also help marketers prioritize their leads and, as HBR states, develop customer lifetime value or make recommendations.

There are several predictive analytics tools in the market. Forrester analyzed 13 vendors in its Q2 2015 Wave report. However, there are a few housekeeping items marketers have to attend to before investing in technology—and continue to upkeep to keep their analysis accurate over time. Here’s a summary of Forrester’s six guidelines featured in the aforementioned Wave report. 

1. Dig for data in multiple places. Professionals need to look for data both internally (across the organization) and externally (e.g. social media, government reports, public records) to help fuel their predictive analytics initiative.

2. Do the prep work. Once the data is acquired, they need to ensure that it’s clean and ready to use. This includes merging multiple sources, getting rid of typos or replicate entries, and filling in any missing information. Granted, developing this set can take time. According to the report, predictive analytics users typically dedicate three-fourths of their time to this type of homework.

3. Create the model. With so many statistical and machine learning algorithms to choose from, predictive analytics users need to narrow their options to find the best solution to achieve their goals. This decision often comes down to data completeness and type of prediction required, Forrester states. The firm also advises professionals to run an analysis on a set of “training data” and allocate some “test data” to measure the model’s success.

4. Consider the effectiveness and accuracy. It’s important to remember that predictive analytics is just that—a prediction. Or as the report puts it, “Predictive analytics is not about absolutes; it is about probabilities.” Still, users want to place their best bets. So, they need to test their established model by seeing if it can predict the “test data” set. If so, they can move on to deployment.

5. Act on the insights. Developing insights is one thing; leveraging them is another. Forrester advises those running the predictive analytics solution to build up trust in the model across the organization and to listen to what their peers find to be the most valuable learnings. 

6. Keep track of the model’s success. As Forrester notes, the quality of the predictive analytics is only as good as the data that’s fueling it. In other words, the model won’t produce accurate results if the data being injected into it is faulty. Users need to ensure that the model remains effective by running data through the algorithms, Forrester states, and adjusting, accordingly.

Even if companies take these precautionary data steps, they can still face challenges. In her blog post, Corey Wainwright, director of content for inbound marketing and sales platform HubSpot, points out a major one—the unknown. No matter how accurate a predictive model may seem, users can’t always account for unforeseen circumstances. Another HBR article also notes that predictive analytics can highlight which areas of the business are doing well and which ones aren’t, which can cause internal tension.

Still, predictive analytics offers a lot of benefits, too. In addition to the aforementioned use cases, a 2015 survey of 150 B2B marketers conducted by Forrester on behalf of predictive marketing software provider EverString found that those who use predictive analytics are nearly three times (2.9x) more likely to report revenue growth at rates higher than the industry average than those who don’t use it. They’re also about two times more likely to be a leader within their industry (2.1x) or consistently exceed goals when tracking marketing’s contribution (1.8x).

So, how many marketers are using predictive analytics and will it continue to catch on? According to the aforementioned Forrester-EverString study, 49% of respondents say they’re using the technology and an additional 40% intend to implement it in the next year.

But just like with predictive analytics itself, the industry can only forecast what the future of predictive analytics will be.

Related Posts