On some long-forgotten dirt-floor showroom, a cave-dwelling salesperson once said, "Let me show you something in a leopard print. It will go splendidly with your beard and that bone you're carrying." Thus was real-time personalization born.
CRM technology has promised to put the power of the "ideal salesperson" to work for the extended enterprise, and personalization is the tool to fulfill that promise. Make the right offer at the right moment, taking into account all that is knowable about the customer's past experience, present situation and future needs, and the world becomes a very appealing place come quarterly report time.
The stakes of personalization have grown up since the days of dot com portal configuration and greeting people by their first names on a follow-up Web visit. According to the Online Consumer Personalization Survey conducted by industry group Personalization Consortium (512 respondents with over two years of Internet experience surveyed in March), customers who respond to properly structured personalization efforts tend to be wealthier than average and spend significantly more money online.
So, the race is on to bring the intuition of that shrewd caveman to every customer contact. The idea is to make better offers at every juncture to keep customers better engaged and more available to future impressions. Making real-time personalization work requires a lot of information and an even greater desire to analyze it and take action.
"The process broadly involves four steps: collecting data on customers, performing an analysis and creating some kind of recommendation to take and then execute that recommendation," says Walter Janowski, research director of the CRM team at Gartner.
While rampant data collection has been under scrutiny by privacy advocates, information is the price of entry for an organization that is serious about personalization, and not just to make accurate predictions about behavior. "The deeper your profile of that customer becomes, the more personalized you can make your interaction, and the more the customer feels he has an attachment to your organization," he says. That loyalty, or stickiness, is the asset everyone is looking to add to the balance sheet. "If the customer goes to your competition and tries to interact, he's building that relationship from square one."
While the most popular notion of real-time personalization revolves around looking for products with particular purchase affinities (burgers and fries, guitars and strings, books and bookmarks, and so on), the refined process looks deeper. If customers reveal a life event that would lend itself to a particular product offering, companies should find a way to make that offer. If a customer is having a consistently bad service experience, perhaps traditional offers should be shelved in favor of an approved retention offer that can save a call to a busy service manager without completely sacrificing the customer's profitability.
If a company's top business priority has nothing to do with a Web site, it may be in luck, since personalization vendors are looking for new outlets for their software anyway. "The pure dot coms went away, so our primary emphasis is on the contact center more than it is on the Web site, by far," says Steve Van Tassel, senior vice president of products for personalization vendor NetPerceptions of Edina, Minn.
Although the Web's instant and accurate flow of data made it an ideal early channel for real-time personalization efforts, pragmatism is dictating the new direction. "The business volume that flows through the contact center is so much higher than the Web site, so it's easy to quantify short-term benefits of personalization," Van Tassel says.
Bell Canada is in the process of expanding a real-time personalization pilot project to hundreds of contact center representatives. The firm used software from E.piphany to analyze customers and cue the agents to present offers most likely to be accepted based on their profile. "Before, it would be a crapshoot, based on what [the reps] liked to sell or what they were comfortable selling," says Michael stanford, group manager of wireless marketing for Bell Canada in Toronto. The company looks not only at basic demographic affinities and product and service offerings, but also profiles customers based on their calling patterns--whether they primarily make or receive calls, when those calls are made, even the average duration of calls--along with third-party psychographic insights.
The end results, stanford says, are more accurate pitches in less time. "There were too many variables for the reps to be able to analyze themselves given the call handling time requirements, so this quickly gives them options with a high likelihood of success."
Since many touches over the Web and the contact center come from non-customers who have little or no history on file with a company, Kevin Cavanaugh, director of product management for CRM vendor Unica, suggests that the personalization process be seeded with as much information as possible from the scant clues available. That might mean inferring a ZIP code and demographic from an automatically scanned phone number or judging a user's interests based on a referring or recently-visited Web site.
The key is to build a profile gradually through a dialogue. "You want to be careful, but when you know a limited amount about an individual, you want to make recommendations that are going to get a reaction," says Jonathan Corr, vice president of product management for CRM vendor Kana. With every accepted--or rejected--offer, "you get them to reveal more about themselves."
At Bell Canada, learning through outright or partial rejection was part of the business process refinements that came with real-time personalization. "Before, if a customer said they were 'not too sure' about an offer, we lost that sales lead," stanford says. But now, a special "customer interested" completion code is entered in the customer's profile, which is passed along to outbound marketing efforts for later (admittedly non-real time) follow-up.
Applying real-time personalization in brick-and-mortar environments is more experimental. An in-store kiosk can deliver the flexibility and availability of a personalized Web experience while taking into account the individual store's unique promotions and inventory levels. Janowski refers to a bank that puts out an in-branch APB to all teller and banker computer screens when a high-value customer has identified herself to an employee, so all of the employees can greet the customer by name and understand the likely purposes for the visit.
Laying Down the Law
Real-time personalization software takes its cues from one of two sources: rules and analytics. As they have for decades, merchandisers can recommend affinities that they see as natural or desirable (the burgers/fries example) and configure e-commerce engines or call center scripts to cue the recommendation when related items are requested by the customer. In other (not necessarily mutually exclusive) cases, a software engine monitors customer activity, compares it against professed likes and dislikes as well as interaction patterns, and through trial and error as well as predictive modeling, begins to build its own set of recommendation criteria.
The complexity of building rules-based real-time personalization tracks grows when the rules are applied to a series of interactions across a complex marketing campaign rather than a point recommendation. "You can define a series of rules, little gates that turn people down one path or the other, and even though you may have [only] ten decision points, the combinations of paths people can take gets very large," says Kim Weins, vice president of marketing for Mountain View, Calif. software vendor Annuncio.
Not only might a rules engine have to tangle with hundreds of overlapping path permutations that could require dozens of uniquely scripted call resolutions, they are also expected to keep track of where customers are with respect to the company's entire slate of marketing campaigns. In the traditional direct-mail world, saturation, if a concern at all, is usually only viewed in terms of the frequency of contacting customers with an offer within a particular campaign or product line.
Such mathematical nightmares lead some experts to stress the importance of software to carry much of the rule-making load. "Today, you're seeing a guess-and-test approach, and if you're doing that going forward you're not going to last very long," says Scott A. Martin, founder and president of the Personalization Consortium industry group. "Marketing and IT have to deploy strong analytics, with a blended approach of declared [customer] segments from the marketing department."
According to Brad Wilson, vice president of product marketing for CRM analytics vendor E.piphany, "With real-time automatic learning models, you do not have to write rules. By sampling the population, the system will automatically learn which offers or outcomes consumers are going to say 'yes' to."
Bell Canada employs both approaches. Prepaid cellular customers are analyzed with a "hands off" model, where the system's self-learning capabilities are allowed to run unfettered to build a matrix of successful offers. For conventional calling plan customers, the company uses a more crafted approach that compares an internally computed sales potential index for each product and each customer to the real-time recommendations made by the E.piphany engine.
Whether using hand-crafted rules or monitoring automatic recommendations from an analytical engine, marketing and product experts need to have their say. The oversight for real-time personalization efforts are "generally business people who understand merchandising and understand what kinds of promotions and offers [the firm] wants to run," says Bill Morrison, product marketing manager for personalization vendor ATG.
Powering the Engine
Because of the ambitious goals, real-time personalization underscores the need for a cohesive, unified customer database that is applied equally competently across the enterprise. "If you're going to do real-time recommendations that are relevant, they had better be consistent with the overall marketing strategy for that customer," Corr says. If high-level customers are put off by seeing general offers or could cannibalize a high-margin offering by being lured to another type of service offered from the same company at the same time, it would behoove the personalization strategy to emphasize identifying such customers before any offers are made, he says.
Yet companies are still struggling with this data consolidation problem, let alone the ability to synchronize across all channels. As Janowski writes in a July Gartner research note, true real-time synchronization is simply not an option for most firms. "While there has been some success with early attempts in, for example, large financial institutions, we believe the technology for maintaining a data warehouse in real time is not yet mature enough for widespread penetration, particularly when collecting multichannel data," he says. Since not all customers require instantaneous awareness across all channels in order to maintain their relationship, however, reaching this pinnacle of sophistication does not have to be top priority. Instead, Gartner recommends that companies conquer the basic challenges of personalization before obsessing over the niceties of instantaneous accuracy.
Because recent times have seen companies first obsessed by growth numbers, then desperate to show short-term profits at any cost, some feel that personalization has been under-applied in being limited to traditional cross/upselling modes. "Right now, most companies that have deployed personalization are still in the early stages, and the easiest opportunity for them to recover the investment in the technology is to increase the size of the immediate transaction--that's cross-selling," Van Tassel says. He believes the long-term "leverageable payoff" will come when real-time personalization is applied to actually reorient each transaction around a customer's implicit preferences, rather than simply pushing out canned offers that seem relevant.
Finding the right business domain to take leadership on a real-time personalization effort--and one that is willing to make the investment in long-term customer satisfaction possibilities--can be difficult because of the ultimate necessity for cross-channel interaction and a sensitivity to the customer's entire profile with the firm. "A lot of companies have trouble identifying who should own this thing since it can apply directly to sales or customer service, and those are sometimes very disconnected operations," says Jim Murphy, AMR Research analyst in customer management strategies.
While real-time personalization is often portrayed as a win-win exchange, it can also be applied to adjust the customer experience in a way that attempts to minimize cost or risk to the organization, by shunting low-value customers away from costly service options, or even rejecting their business altogether. "Financial industries are very interested in fraud detection. If you have predicted that a customer is likely to commit fraud, is it appropriate to discourage the customer from transacting?" Janowski says. "It's a fair question whether that's ethical or not."
Also, more accurate real-time judgments about the character of a customer lead to a greater opportunity for the fairly unpopular (among consumers) notion of dynamic pricing. Annuncio's Weins cautions companies to think twice before heading down that road even if customer sentiment is not a deterrent. "You can say there's a big negative to the seller, because people believe everything is negotiable," she says. A known dynamic pricing structure could encourage customers to mask or change their profile based on browser settings, or be disinclined to call their designated contact center for service. "People will see if they can get a better price by coming back later or going to a different site, so it can be as much a negative implication as the positive from finding people who may be willing to pay a little bit more," she says.
Unica's Cavanaugh warns firms to maintain a healthy suspicion about customer-reported data and to treat it as only one of the many components in an overall personalization profile. According to Personalization Consortium's research, 33 percent of e-commerce buyers have deliberately misreported personal information. "It's important to recognize that not everyone wants this kind of attention, but the vast
majority just want good service," Morrison says.
Rolling With the Changes
If real-time personalization is truly going to become a contributor to customer satisfaction and retention, it may make sense to think of it as a tool to guide and funnel a company's natural instincts to pitch endless cross-sell offers in the hopes of finding a few that stick. "It's more about what not to provide to people than what to recommend," Martin says. "Take away the 'annoying' factor, don't ask me the same thing twice. I think that's a big oversight [among real-time personalization adopters]. They look so much at cross-sell and upsell and not how to close the gap between consumer compromise and annoyance factors."
Indeed, being asked to buy two things when a customer only came for one is not necessarily greeted with undying adoration. "Sure, the consumer is 'better off' if a cross-sell recommendation is one they like rather than one they don't, but it's not a visible, changing-the-nature-of-the-interaction kind of thing," Van Tassel says. "The real promise from the consumer standpoint is still coming down the road," when real-time personalization is used as a retention and reward tool and enables companies to streamline not just recommendations, but the entire sales and support process around an individual's preferences.