BtoB

Predictive analytics

By Published on . 0

Reprints Reprints

How to define predictive analytics? Sometimes the term is conflated with more traditional analytics, like CRM applications that help ballpark nurture-worthy potential customers or dashboards that give clear snapshots of campaigns metrics to prompt future changes of course. But these offer simplistic analyses compared with what predictive analytics can provide.

"The big challenge is how to access data that are relevant to a potential decision," said James Thomas, VP-business intelligence tools at business intelligence software company Business Objects. "Data can either validate a decision that's made or recommend scenarios."

A key characteristic of predictive analytics, said Thomas, is the ability to pinpoint anomalies, so-called "outliers," that can signal greater or lesser opportunities, uncover untapped potential, ID customers at risk of leaving and signal where to reallocate resources.

"You might think of a simulation of what this can do," said Colin Shearer, senior VP-market strategy at leading predictive analytics company SPSS Inc. "If you were able to do your marketing, for example, to those with the highest propensity to buy, you can actually hit 85% of the buyers in the first 30% of the population."

Business Objects this month teamed up with SPSS to meld the SPSS forecasting tool to Business Objects' data mining abilities with a simple user interface. Business Objects, which has used SPSS for its own marketing analytics, will resell the enhanced capabilities to its own business intelligence clients.

While predictive analytics would seem to have its greatest potential in b-to-c, with its massive customer databases, b-to-b marketers "have more transactional information, rather than customer descriptive information, so it balances out well," Thomas said.

CheckFree, a provider of electronic bill-paying services to financial institutions, has used predictive analytics in trouble-shooting its services, where snafus in the service-delivery process are the anomalies Thomas refers to. CheckFree has the advantage of tapping into both transactional information (its bank clients) and customer data (the bank's own massive list of clients).

"Electronic billing is in its infancy, so we developed a project to ensure we get optimum bill delivery, that we are not missing any bills," said Jane Damschroder, senior business analyst with CheckFree. If the metrics alert CheckFree that it hasn't gotten a bill when expected, it can go back to the banking merchant and see where the failure was.

CheckFree's predictive tool—called JMP, offered by SAS and implemented in April 2007—identifies the company's own version of outliers, in this case isolated troubles that may lead to client attrition.

"We want customer satisfaction because, if your bill doesn't arrive, you won't be satisfied with your bank's service and you won't sign up for any other electronic bills," she said. CheckFree statistics also indicate satisfied customers are more likely to recommend the service to a friend and less likely to switch banks.

"We've seen significant improvement here," Damschroder said, noting that the quarter ended Sept. 30 saw a jump of 6% in e-bills sent out, at a current rate of 63.9 million a quarter.

While predictive analytics is valuable in parsing today's massive amounts of data, forecasting also can be prompted by traditional descriptive analytics that occur throughout the sales cycle.

"Most marketers don't have the big bucks for the big predictive-analytics suites," said Tom Judge, VP-strategy and business development at Direct Marketing Partners. "They have to do tactical marketing, so it's more useful for them to develop a series of benchmarks in market subsegments, collect information about customers' buying patterns, analyze them weekly and make quick decisions that can change marketing variables," he said.

One of Judge's marketing clients agrees with this simpler approach, using simple Excel spreadsheet charts.

"You start to see quickly how receptive a target is to a particular message, and whether you're getting through to a decision-maker," said Jennifer Walsh, senior director of marketing at brand-protection company MarkMonitor Inc. "And you can dig down further for more vertical reactions, to see how they differ from the market at large. In a few weeks, you begin to spot trends."

But if anything characterizes predictive analytics versus, say, simple spreadsheet workups, it's the automated aspect of its forecasting.

"The question is, when should you tweak?" said Jon Weisz, director of marketing at SAS. He said that traditional descriptive analytics, even attractively displayed by dashboards, don't make recommendations about actions. "You don't want to be a `wiggle-watcher,' taking action when nothing's substantively changed."

Predictive analytics, Weisz said, provides a "smoothing" ability, helping eliminate wiggle-watching by identifying trends that are real and more actionable. Then the tool has the ability to take that action itself.

"People are very bad at estimating and we tend to overreact," he said. "More advanced companies are trying to use predictive analytics to ask if the trends they're seeing are real."

In this article:

Read These Next

Comments (0)