Marketing’s Holy Grail: Digital personalization at scale
November 22, 2016
By Brian Gregg, Hussein Kalaoui, Joel Maynes, and Gustavo Schuler
Personalization drives growth. But to scale it, companies need to do four things.
Customers decide very quickly—in a matter of seconds—whether they like your marketing message. Provide something relevant and you’ve got a satisfied customer. Miss the mark, however, and they’re gone.
This issue of relevance in our era of instant gratification is particularly pronounced because consumers are bombarded with messages, most of which are off target. Personalization—the tailoring of messages or offers to individuals based on their actual behavior—promises to address this issue.
While many companies have been able to personalize with a few product lines or segments, most still struggle to scale across all the ways they engage with customers. And although technology has an important role to play, in our experience, most companies already have plenty of tools. The real challenge is to transform the marketing organization’s processes and practices to achieve the full potential of personalization.
Done right, personalization enhances customers’ lives and increases engagement and loyalty by delivering messages that are tuned to and even anticipate what customers really want. These benefits to the customer translate into benefits for the company as well. Personalization can reduce acquisition costs by as much as 50 percent, lift revenues by 5 to 15 percent, and increase the efficiency of marketing spend by 10 to 30 percent.
Through more than a hundred engagements over the past five years, we’ve found four steps that lead to successful digital personalization at scale:
Step 1: Take a journey lens: Use behavioral data to find where the value is
The foundation of personalization is acting on behavioral data. The first step is to group customers with similar behaviors and needs. For example, mothers who exclusively shop a brand for their children or fashion-conscious young women who buy new private-label styles. Most companies find it useful to start with eight to ten such behaviorally based segments as a first step in their evolution to 1:1 marketing.
The next task is to understand, for each segment, the customer journey —the series of interactions with a brand from initial consideration, to purchase and use, and then to subsequent purchases. Marketers can do this by integrating information from internal sources such as visits to the company website, purchases at a store, or calls to the contact center, with information that can be acquired from external sources, such as prospects’ visits to a competitor’s website.
Combining these segments and customer journeys creates hundreds if not thousands of “microsegments,” which form the basis of 1:1 personalization. Not all microsegments are created equal, of course. The potential of each must be evaluated and prioritized carefully, based on relative value. For example, a leading retailer we worked with determined that it is more valuable to engage its customers within their “resupply” window—e.g., by reminding them they may be running out of toothpaste and their favorite brand has a limited-time offer—than by pushing them to go deeper in the category by suggesting other oral-care products such as mouthwash or teeth whiteners.
Step 2: Listen and respond: Plan in advance to react quickly to customer signals
Personalized marketing is a two-way street: The customer provides signals—information about his or her needs and intentions—through activities like purchases, online browsing, and social media posts. The company responds to the signal with a relevant and timely message, which we call a trigger, that is sent to the individual customer.
Doing this effectively requires careful advance planning. The marketing team needs to develop a library of trigger messages matched to individual signals.
Trigger messages can be of different types—images, copy, titles, offers—that can be combined dynamically to match the situation.
Coming up with trigger events involves creative problem solving grounded in sound analytics. For example, a next-product-to-buy algorithm based on machine learning could send a message suggesting a set of related products triggered when mothers have clicked on a different product but not bought it (see sidebar, “How personalization that works creates value for customers”).
Mary is a mother with two children in primary school. Early last August, she visited one of the retailer’s stores to buy items for her kids, including several she’d previously viewed online. Those items were logged and attached to Mary’s profile in the store’s database.
This July, Mary browses children’s clothes on the same retailer’s website but does not click “buy.” Her combined offline + online behavior during the back-to-school shopping period as well as the items in which she’s shown interest in create the signal that Mary may be open to making her first back-to-school online purchase for her young children.
Within 24 hours of her most recent browsing action, she receives a personalized email offering a 10 percent discount on select items she’s been reviewing, if she purchases them online. The message explains that the process is safe and easy and includes tips on how to move selections into the shopping cart and check out. It also suggests additional items she might consider, based on analysis of her past purchases and recent browsing behavior.
For all the preparation, getting the full value from triggers requires a test-and-learn process: sending an initial message, evaluating the results, altering the trigger, and measuring the results again. It typically takes four to five attempts to refine a personalized trigger to capture 80 percent of its potential value. For example, a leading apparel retailer we worked with went through four different iterations of a next-product-to-buy email until it found the winning formula, which ended up yielding twice the impact of the first iteration. Additional refinement after that usually yielded diminishing returns.
These sorts of personalized triggers have been shown to be three to four times more effective than blast messages. They can also introduce the customer to new products and new modes of interaction (such as buying online versus in store) that are more convenient, further enhancing their experience.
Once a signal and associated trigger have been shown to be valid and refined, it becomes a business rule, and all future customers associated with the signal automatically receive the appropriate trigger message.
Step 3: Build the war room: Empower a small group of the right people
Going from the traditional marketing calendar to personalized triggers sent in response to individual customer’s signals means shifting to a radically different way of working. This is where we see many organizations get stuck. The secret to kick-starting this change is to empower a small group of the right people. To quote cultural anthropologist Margaret Mead: “Never doubt that a small group of thoughtful, committed citizens can change the world; indeed, it’s the only thing that ever has.” The same can be said of your company.
We call these small groups “war-room teams.” They are staffed with 8–15 carefully selected people, including a campaign manager and staff from creative, digital media, analytics, operations, and IT. A medium-sized e-commerce company might need just one such team; we have seen large organizations with as many as eight teams like this working at once. To be successful, the war-room teams need executive sponsorship at the very top of the organization to remove roadblocks and empower them to get things done, often through creative workarounds.
This is not a task force where people come together a few hours a week while staying in their prior jobs; the team members must be dedicated full-time to the war room. Their job is to drive business results—not merely clickthroughs or page views, but a materially better customer experience and actual dollars. At a wireless carrier, for example, the objective might be lower churn for multiline households; at a retailer, increased follow-on sales to new customers in their first 90 days by introducing them to more-convenient online purchases or products they did not know about.
A war-room team does this by day-in and day-out searching for the signals with the highest potential value and developing, launching, and iterating on personalized messages that get results. At the start of a personalization program, each team should launch one to two new triggers per week. As they gain experience, they should be launching one to two new triggers per day.
Step 4: Rewire and hardwire: Focus on the processes and technology that really help teams work faster
To work at this pace at scale, agile processes have to replace the old ones. For example, in the new world where copy, creative content, and templates can be quickly mixed and matched to create a near-infinite number of personalized variations, cumbersome and lengthy approval processes simply have to go. Teams need to be empowered to act on their own to quickly test and iterate different ideas and zero in on what works. Mistakes will happen, but leadership needs to be ready to accept that, learn lessons, and move on.
The second thing needed to work at this pace at scale is the right automation technology. Assembling and operating a marketing-automation tech stack, however, can be a challenge.4 There are literally thousands of providers of tools, and despite vendors’ claims, no single party offers a true end-to-end solution. Companies are left on their own to filter the hype, figure out what to buy, and do the integration themselves. The result is all too often automation software spitting out millions of messages that amount to little more than spam.
The technology’s job should be to help to the war room teams find signals and deliver triggers more efficiently. So the work of these teams should guide the evolution of the tech stack. Proceeding in this way, the organization can set the right requirements and parameters to help narrow in on the tech stack solutions that will create real value.
Marketers often view personalization at scale as a daunting undertaking, requiring millions in IT investments. But successful players often start small, generate top-line impact quickly—in a matter of weeks, often—and self-fund the initiative after that. Only then can you invest in the automation and institutionalization of new ways of working across the whole organization over time.
About the author(s)
Brian Gregg is a partner in the San Francisco office and leader of McKinsey’s Digital Marketing Practice; Hussein Kalaoui, Joel Maynes, and Gustavo Schuler are associate partners in McKinsey’s Los Angeles office.