The answers to three simple questions all marketers need the answers to—“What will be?”; “How will we get there?”; and “What is it worth?”—are locked up somewhere in your organization’s data warehouse. The disciplines of predictive analytics, marketing mix modeling, and customer lifetime value analysis are the art and science by which those questions are answered.
Marketers who feel they’re doing everything they can with their data analysis tools should consider the tremendous opportunity that idleness opens for competitors willing to take advantage of the dramatic gains in data storage and computational power made over just the past few years. “Today, you can run a report on 100 million people and have an answer in minutes. Five years ago that would have taken a day,” says David A. Steinberg, CEO of XL Marketing.
Analytics give marketers the strategic advantage not only to beat competitors to the punch, but also to anticipate customer demand before it is even felt. And interest in analytics is growing. Gartner Inc. projects a $13.8 billion worldwide market for analytical and business intelligence solutions this year, a 7% increase over 2012. The ability to simply collect customer data is no longer a competitive differentiator—brands must have the tools and expertise necessary to gain insights from that data. But first, they must know which questions they want answered.
What will be: predictive analytics
Knowing what happened yesterday in last week’s campaign is table stakes in the modern marketing world. Predictive analytics provides insights—or at least very, very good guesses—about next week’s results from tomorrow’s campaign. It does this by using a wide range of historical, demographic, and behavioral data points and trigger events to model everything from response rates to complaints, and is an important component in shifting marketers from spray-and-pray methods to social, collaborative strategies. “In the past 10 years our clients have changed from trying to convince people to buy what’s in their inventory, to trying to understand what customers want,” says Erick Brethenoux, director of business analytics and decision management strategy at IBM.
Brethenoux credits social media with changing how brands view the value of predictive analytics, because peer recommendations completely out of a brand’s control can make or break even the most elaborate marketing campaign and predictive model. He also points to its role as a key input in today’s predictive models. “To make good predictions, you need intent data, and with social media people are telling you what they’re going to buy and when they’re going to buy,” Brethenoux says.
Predictive analytics doesn’t have to focus solely on people, however; it can be applied to products, as well. Analytics provider Target Data built its flat list of residential homes for sale into a powerful predictive tool that understands more about property than owners do. “We realized that we knew when homes came on the market and when they came off, and from there we should be able to predict how long they are going to stay on the market,” says Scott Bailey, Target Data’s EVP of strategy and analytics. “A lot of advertisers are more interested in the back end of that window, and want to be in front of people when the home is going to come off the market—whether the owners know it or not.”
Of course, intent to stop buying can be even more important than intent to buy. For instance, telecommunications providers XO Communications five years ago faced higher-than-expected churn rates. The company began an extensive campaign to improve small business retention, with the goal of identifying and intercepting signals that predicted a high probability of contract non-renewal. The traditional approach of reaching out with a phone call or mailer as a contract approached the renewal date was clearly not working to keep churn within an acceptable band. “We knew we couldn’t do the same thing over and over, but we didn’t know if we should throw more people at the problem, talk to our customers more, or do more offers,” says Bill Helmrath, director of business intelligence at XO Communications.
Working with IBM, XO began by attacking its qualitative biases about the reasons customers churn and decided to build detailed, data-driven profiles of churn risk. The 25 variables range from obvious factors, such as the frequency of service outages, to more esoteric criteria such as whether the customer is in XO’s home network. With these variables in hand, XO assigned a churn risk score on a scale from one (likely to renew) to 10 (highest risk of churn in the next 120 days).
XO made the strategic decision to consider no customer a lost cause and has focused the lion’s share of its retention efforts on customers with the highest likelihood of churn, with offers that the predictive model indicates will best address their concerns. For instance, customers on outdated, higher-priced service packages will be offered more competitive 2013 pricing, while customers with complex phone networks may be offered an automatic call routing service. The predictive analytics project also brought the company’s marketing and customer care group in to collaborate to identify key customers and devise retention strategies. “Before, this would have been handled by customer care alone. Now we’re working together to find the best plans to save customers,” Helmrath says.
The company-wide collaboration and focus on the highest-risk customers has paid off handsomely for XO, to the tune of a 26% reduction in churn rate and $3.8 million annual ROI.
How will we get there: marketing mix modeling
At its most basic, marketing mix modeling attempts to identify the most effective combination of channel and message for a particular business goal, including acquisition, outreach, retention, and upsell or cross-sell campaigns. Many organizations use it to take a longitudinal look at campaigns to identify the winning strategies over time, and attribute marketing successes and shortfalls at a granular level. Getting serious about marketing mix modeling is an important step in becoming more strategic in marketing execution because it involves identifying the true cross-channel costs and ROI associated with broad business goals, such as greater prospect conversion or reduced customer attrition. Ultimately, marketing mix modeling provides better understanding of marketing ROI on a channel-by-channel basis. It shows the results of these cross-channel interactions.
Though not a true Big Data discipline, marketing mix modeling does demand a commitment to a large amount of data. Unifying customer interactions across social, digital, offline, and point-of-sale channels is not a trivial task, particularly when customer behaviors, such as clickstream activity and purchases, must be merged with spending and outreach activities. Success usually requires refined and specific customer segmentation.
Consequently, consumer brands have generally led the way in this discipline, with B2B counterparts working to catch up. “The B2B side of [mix modeling] is still pretty new—if you ask most B2B marketers to define their segments, they give you a combination of the prospect’s industry and how big they are,” says Jim Swift, CEO of Cortera. “There needs to be more variables in the matrix that include the characteristics of the people you contact, because you could be targeting the right business but the wrong area and getting no results.”
The challenge of mix modeling is even greater in markets with complex purchase processes. Like most drug makers, Astellas Pharmaceuticals U.S. must consider the impact of its marketing on a variety of customers, including patients, physicians, pharmacies, managed care groups and even the federal government. As a result, the midsize developer of biopharmaceutical treatments faces a tremendous set of challenges in modeling its marketing activities. “We need to figure out which tactics drive conversations between doctors and patients, we have to know how the frontline sales force interacts with physicians, make sure our products are stocked in pharmacies, and make sure insurance companies allow patients to receive the drugs,” says Chad Dau, associate director of marketing analytics at Astellas. “But the technology has come a long way, and we can do analyses we couldn’t attempt 10 years ago.”
The mix models have led to a number of strategic changes at Astellas. Several educational programs were redesigned when the model detected that they were not promoting the expected levels of patient and doctor uptake. Performance testing is now done on demand-generating activities for each of the populations—such as doctors, pharmacists, and end-user patients—if that makes up the demand for Astellas’ products. Inefficient spending is culled, and new spending is tested to look for an ROI lift of 5% or more. Most significantly, Astellas found by examining response rates and conducting control-and-experiment group testing that it could greatly reduce the marketing spend on its oldest and best-established products without affecting sales. “Everybody knew we could pull back on spending for the mature brands, but without hard information it was difficult to do,” Dau says. “The model data gave us the backup we needed to do so.”
Unlike many of its competitors, Astellas does all of its marketing modeling in house, which Dau believes is a competitive advantage as it provides more accurate and honest insights. “Most companies outsource this process, but we implemented SAS in house and brought in a talented team,” he says. “Working in house lets us use multiple methodologies and try a lot of different things. External vendors just tell you ‘Everything works!’”
What is it worth: customer lifetime value
Customer lifetime value (CLTV) analysis attempts to model the total revenue or profit stream from a customer over the entire duration of the relationship. Many organizations consider it a predictive process, because it attempts to model future purchases, often mapping out the anticipated dates of purchase. These models often factor in inflation rates—for instance that a $100 purchase in one year is not as valuable as a $100 purchase today. CLTV is not a fixed, immutable figure—successful campaigns can increase lifetime value by encouraging more or sooner purchases, while negative brand experiences should be included to reduce the expected income.
Enhancing CLTV can start with simple, easily stated goals, such as finding effective ways to rekindle a relationship with a dormant customer. WWRD U.S., best known for its Waterford brand of crystal products, began a complex campaign in June 2012 with a basic mission statement: Bring back more direct customers for repeat purchases, with a particular emphasis on luring back one- and two-time purchasers. This consumer segment represents a much warmer opportunity than a new customer acquisition campaign, and one closer to an eventual pattern of loyalty. “As a luxury manufacturer and retailer, the economy affects us, so we need to find relevant reasons for our customers to buy,” says Joseph Schmidt, director of e-commerce in the Americas region for WWRD. “We don’t want people going 366 days without making a purchase.”
The majority of Waterford customers are one-time buyers, giving Schmidt and his team a rich vein of prospects. Working with email marketing solution provider Listrak, WWRD developed a lifecycle grid for all of its buyers, sorted by number of previous purchases and days since last purchase.
Each customer segment represented in the lifecycle grid received a different message. As a luxury brand, WWRD is not interested in discounts. Instead, Waterford buyers received offers based on gift wrapping, engraving, shipping, and even simple lifecycle messages thanking them for their purchases and reminding them of upcoming gift-giving opportunities. WWRD altered the offer type based on the number of purchases each customer had made. Additionally, the time since the last purchase affected the offer creative. WWRD used the Listrak system to input different offers and optimized those offers on the fly based on response rates.
The brand paid special attention to balance the business drive to increase customer value against the comfort of the customer, to ensure that messaging did not unnerve the audience. “We paid a lot of attention to making the messages not seem creepy. Saying ‘We see it has been six months since your last purchase’ was too Big Brother-like,” Schmidt says. “Instead, our messages were more about recognizing people as ‘one of our most loyal customers’ or ‘as someone who has shopped with us a few times.’”
The results are significant. WWRD customers receiving lifecycle offers converted 28% more often than those in the control group, and 220% more often than all website visitors. The greatest gains in lifetime value came from customers who had made just one or two previous purchases—fully 80% of revenue from lifecycle offers during the holiday 2012 season were from the customers with the shortest purchase histories, and most of that was from one- or two-time buyers who last purchased over one year ago.
“At this point, we know we’re capable of getting people to buy again,” Schmidt explains. “With a win under our belt, the next step is to look at whether these customers are buying for themselves or giving gifts, and whether those lifetime values are different. That data is the next to come.”
Lifetime value also extends to brands with a less-tangible product than a crystal vase. Heifer International, an antipoverty and hunger charity, deals in gifts rather than orders and fulfillment, but in many ways it operates like any other direct marketing organization. Consciously or not, Heifer’s potential donors hold it to the same standard they would a retail brand. “The world is full of customized feeds and offers, but the typical nonprofit has been shipping out the same message and hoping people will respond,” says Ashley Michael, email marketing manager at Heifer International. “Over the past seven years we have been trying to move in the direction of the for-profit world with customized messages that are timely and connect on an emotional level.”
At Heifer International, maximizing the lifetime value of a donor means converting the occasional or annual giver into a “sustainer”—someone who gives in a monthly and automated fashion. Heifer made the strategic decision to focus on increasing donation frequency, rather than average annual value, believing that the loyalty and connection of a regular gift provides the greatest overall benefit. “Our goal is not just a second or third gift. We want to have a relationship and build brand loyalty, and have our donors connect to our mission in a very personal way,” Michael says.
Sustainers provide high lifetime value, but also have higher acquisition costs. So Heifer has worked with database marketing company Acxiom to model the most likely existing donors who might convert to sustainers. The real-world results reflect a work in progress. The test campaign produced a 26% boost in average gifts among the recipients, but a response rate 23% lower than the control group. “Obviously, it’s ideal to have both a higher response rate and higher gifts,” she says. “This emphasizes that modeling requires iterative testing to strike the perfect balance.”
Just as Heifer found the need to continue to refine campaigns and consider the significance of its finds, brands with a serious commitment to analytics know that there is no finite destination for analytical work. If anything, demand from business leaders for analytical insights is only going to grow. “Attention spans are getting shorter, and people want results in an hour,” Dau says. “The appetite for the kind of information analytics can provide to brand teams has come a long way.”
