CRM + Predictive Analytics: Why It All Adds Up
Though predictive analytics (PA) tools have been around for decades—with a strong uptake historically in telecommunications and banking—demand has risen dramatically in the past couple of years. In fact, any CRM vendor focusing on a business-to-customer-facing client base must incorporate some PA into its offerings in order to build staying power.
The biggest factor driving demand for PA in recent years is return on investment (ROI). Since the onset of the recession—which may or may not be over, depending on your industry and location—many businesses have squeezed more value out of every dollar spent. With layoffs, furloughs, and closings, every aspect of an enterprise’s budget has come under scrutiny, and CRM is no exception. Consequently, in those kinds of evaluations, PA’s promises of targeted and optimized customer outreach have been attractive.
Traditionally much of the ROI derived from PA involves “maximizing the lifetime value of a customer,” which in many cases refers to customer retention. That means intervening with a next-best offer when a customer appears likely to turn away from a provider or making the right offer once the customer has announced his intention to break ties. In some cases, it’s vital to determine when a customer seems likely to leave before he says anything.
“If people actually tell you they’re going to leave, it’s much more difficult to retain them,” says Rob Walker, vice president of decision management and analytics for Pegasystems, a business process management (BPM) and CRM solutions provider.
CALCULATING LIKELY DEFECTIONS
To make that determination, PA solutions leverage weighted algorithms and models, sometimes in the hundreds, simultaneously. For example, a telecom company would employ PA to figure out when customers are likely to leave, incorporating a number of factors.
Among them are trigger dates, such as the expiration of a contract, as well as call logs and wireless browsing history. PA would notice a customer who has called the service line of a competitor or looked on its Web site. Also important is usage: If a customer already has signed up with a competitor, maybe he’s just using the time left on his previous carrier’s phone. Factors such as those get processed through a PA tool to assess how likely a customer is to leave. If the returned value indicates that flight is likely, an intervention can be made.
Exactly what should be done is another determination made through PA. The expected future value of a customer’s business is weighed against how much a given offer for retention would cost, and that is stacked against a customer’s likelihood to accept it. Of course, other considerations are analyzed. A PA model may take Bureau of Labor and Statistics information, for instance. If a customer lives in an area of high unemployment and falls within certain demographic lines, he might not be able to afford his current plan. If he’s been laid off, a cheaper plan should be offered. A plethora of determinations like these must be made.
In any case, the PA tool will process all the information and produce a next-best offer, in real time. If a customer says he plans to leave, an agent can punch that in and the PA can give the agent a counteroffer to try to retain that customer.
This approach can be effective. Pegasystems had one client, a top three British telecommunications carrier, Orange U.K. (which is now a part of T-Mobile U.K.), that retained an additional 4 percent of its most valuable customers each month, which added a gross operating profit of nearly $40 million per year after it implemented Pegasystems’ PA solution, according to Walker.
“[And] this was actually during the implementation,” Walker says. “I don’t think all of the agents were even enabled.”
To put that into perspective, Orange has more than 17 million mobile and broadband customers. Scaled to a larger business, Walker suggests, it could swell to $100 million a year. A princely sum if Orange’s claimed returns carry through.
There are, of course, cases like this one from other BPM and PA vendors. SAS says in a release that one of its clients, 1-800-Flowers (a U.S.-based flower delivery company with almost $1 billion in annual revenue), saw customer retention jump by 10 percent during the recession. That translated into $40 million in additional revenue. The company was still down $98.4 million in net income in 2009, but PA can help businesses land more gracefully after a downturn-driven stumble.
GAMBLE PAYS OFF
Harrah’s Entertainment, a gaming conglomerate that counts Caesars Palace among its more than 40 casinos worldwide, provides a case in point. The company has been using PA tools from SAS for more than five years that work with reporting solutions from Cognos and campaign management tools from Unica.
Harrah’s system contends with 10 million customers and a tremendous amount of transactional data from its casino floors everywhere, from Las Vegas, Nevada, to Biloxi, Alabama. It’s all part of the company’s Total Rewards program, which tracks customers through Harrah’s resorts, restaurants, and gambling houses. The tools produce actionable items for floor workers in real time by optimizing next-best offers to ensure that a customer staying or gaming in a Harrah’s property wants to spend most of his vacation within Harrah’s holdings and would want to return. Essentially, the PA drives Harrah’s marketing and operational decisions.
In 2003, combined revenue at the company’s two Las Vegas properties rose by 10 percent, and income from operations jumped by 26.6 percent, according to an SAS news release. The same release says that gaming by customers using Total Rewards cards increased by 5.6 percent, and gaming revenue generated by customers outside their home markets climbed by 15.6 percent.
But the current economy paints a different picture, says David Norton, the chief marketing officer for Harrah’s. “The gaming industry has been hit hard by the economy,” he says. “So our revenues are down over the last couple of years because gaming is definitely more discretionary. People aren’t spending their money. Our company used to be closer to $10 billion, and now we’re closer to $8 billion.”
He adds, however, that Harrah’s nonetheless has managed to outperform its competitors in a constricting market, largely guided by PA tools.
“Really it’s about understanding which customers are performing worse than others,” Norton says. “Is it people that live farther away or closer? Is it younger or older? Slots versus tables? That ability to understand the issue at a very refined level meant we could target customers whose business with us had declined, as opposed to doing more broad-brush activity.”
CLUES IN SOCIAL MEDIA
Also contributing to the spike of interest in PA is the rise of actionable social media data from external consumer communities. Networks like Facebook and Twitter have managed to net an enormous sea of users—far more than any opt-in network a business might hope to build for itself. The potential to learn about current customers is enormous, as is the ability to define and target prospective customers. There’s been no shortage of heralds belting out “the future is social CRM” through any trumpet within reach.
In some cases, enterprises have tapped those networks to gain deeper insight into customer behavior. Social media has been used to support market research and development; to get out marketing messages faster; to define segments of an enterprise’s consumer base; and, in some cases, to promote self-service as a way to drive down support costs. The real show-stopping stuff is yet to come, though.
“Over the next several years…[expect] a shift from stand-alone communities to external socials integrated with CRM systems,” writes James Kobelius, a senior analyst for Forrester Research, in an email to CRM magazine.
Moving forward, Kobelius foresees these external communities becoming an important channel in an enterprise’s overall CRM initiatives—almost like a third column call center or Web portal in which “super-customers” (hopefully incentivized) act as conversation facilitators and troubleshooters. Some of this is happening already, but Forrester sees it growing. These company superfans basically would act as outsourced agents.
As Kobelius puts it, “Customers feel…each other’s pain more acutely than almost any service rep can possibly pretend to. Just as important, longtime, passionate customers are often much better informed than somebody hired off the street and rushed through basic product-support training.”
If this vision comes to pass, the implications for data gathering and the application of predictive analytics are numerous. Information from social network profiles, posts, click histories, and usage logs could help define customers by identifying influencers and leaders within social groups, as well as their followers and outliers. In essence, they could be used to discover links among people, organizations, and businesses that would otherwise escape the attention of users.
“Social network analysis can generate a fine-grained, multifaceted picture of patterns of cooperation and collusion, coalition and co-dependency, influence and deference, and affiliation and isolation among and within groups, within the context,” explains a recent Kobelius report titled “Zero In On CRM HEROes: The Role of Social Network Analysis.”
The marketing potential is clear. Apart from knowing one’s customers, a company can gain tremendous insight into customer demographics outside of its base. Enterprises have access to their customers’ interests, passions, friends, family, and even minute-by-minute thoughts in some overzealously update-conscious cases. The demographics picture is much more detailed than it has ever been. And all of this information doesn’t run through the filter of an in-house study or survey. Using PA, these factors can be leveraged against almost any business decision—from the price of a widget to the color of its packaging.
PA’s value in the social realm may also prove worthwhile in customer retention, which is the historical value of PA. Dissatisfied customers who are likely to turn from a business can be discovered over social networks, or at least contextualized. If they have relationships with other users, it would be possible to build offers around those relationships. Forrester’s super-customers might even be an appropriate—and cheaper—way to apply a company salve to an irritated customer. In some cases, a company could simply have a fellow customer reach out to someone who’s had a bad customer experience.
Given all that, it’s not hard to see why businesses would be keen to develop robust PA. Interest has generally percolated initially from public relations departments, perhaps looking to avoid the next social media disaster, like “United Breaks Guitars.” But, ultimately, the usefulness of PA in social networking is so compelling that almost every department in a company has a stake in it.
“I have to say social media is a lot like retail sales data—it isn’t owned by any one group,” says John Bastone, global product marketing manager for SAS customer intelligence solutions.
Every department of a business can find use for greater intelligence about who their company’s customers are and what would be the most efficient ways to reach them, he explains. Given all that, it’s no wonder Walker brags that PA’s business proposition is “so easy to sell it’s ridiculous.”
GETTING ACQUISITIVE
Vendors seem to know it, too. BPM companies without strong PA offerings are looking to forge partners or make acquisitions.
Toward the end of 2009, IBM acquired SPSS, a major PA vendor. After an intense two weeks of negotiation, SPSS sold for $1.2 billion, which reveals how much IBM values PA technology in its own future business. Korbelius sees this as an important move. For IBM, it means strong integration of a hotly desired tool; for SPSS, the acquisition was “essential for it to hold its own against SAS.”
On a smaller scale, 2010 saw Pegasystems buy Chordiant, another PA provider with a solid track record, for $161.5 million, or $5 a share. And Sword Ciboodle, a customer interaction software provider, partnered with SAS to integrate PA.
This trend will likely continue. CRM companies that do not embed advanced/predictive analytics, Kobelius writes, probably will partner with SAS, IBM/SPSS, KXEN, and other PA/[data management] vendors. Or, they will become key acquisition targets for vendors of products primarily deployed for customer-facing business processes, such as business intelligence, analytics, data warehousing, marketing campaign optimization, and social media analytics. Among those companies likely to be pursued, he cites CDC, FrontRange, Maximizer, NetSuite, RightNow, Sage, Salesforce.com, and SugarCRM.
Eric Barkin is a freelance writer based in New York. Check out his blog at ericfelipebarkin.wordpress.com.