Marketing has entered its “uncanny valley” moment. That term was first used to describe the deep unease people feel when a robot or computer-generated human facsimile nearly succeeds in passing itself off as the real deal. It may walk like a person and talk like a person, but without the nuance, context, and tone that make for real engagement, what’s left rings false. And that’s off-putting.
Some marketing efforts give off the same vibe, like that creepy feeling when a casual online search leads to a glut of ads for the same pair of boots or getaway destination. While this type of engagement represents a step forward in personalizing experiences with customers, clumsy efforts at retargeting often feel intrusive and annoying.
It’s worth trying to get right. We know that personalization can deliver five to eight times the ROI on marketing spend, and can lift sales by 10% or more. Although the marketing industry has been promising personalization at scale for the past 20 years, expecting a machine to generate the perfect personalized experience is a fool’s errand. Rather, we’ve found the best way to achieve meaningful personalization is by systematically testing ideas with real customers, then rapidly iterating. Until recently, however, the tools and capabilities to execute this operation—delivering truly relevant personalized offers and content to millions of customers and prospects, across channels, content formats and touchpoints–have not existed.
That’s finally changing. Marketing technology, automation, and advanced analytics techniques have now reached the level where effective personalization at scale is possible. And yet getting this “test and learn” engine to run properly requires a fundamental re-architecting of a company’s marketing analytics processes. The goal is to create a learning ecosystem, one that connects insights to outcomes as part of a continuous, self-improving cycle.
Integrating the Three D’s
This requires the integration of three things: data discovery, automated decision making, and content distribution.
Data discovery is about sourcing and combining traditional and behavioral data to uncover meaningful insights about customers (such as their preferences, interests, and needs.). Not that this is easily done. Given the complexity of coaxing meaning from a wide range of data, companies tend to limit the data they use, generally focusing on the data that’s easiest to get. In addition, traditional CRM systems, built on more rigid, relational databases, often don’t have the flexibility or scalability to manage vast piles of structured and unstructured data. What companies need are systems that can run the advanced analytics to discover useful and practical insights, and then trigger the sending of appropriate messaging, e.g., if customer “A” does action “B,” send item “C.”
An emerging answer to this issue is the customer data platform (CDP), which is the modern version of a customer data warehouse—though one that is far more flexible and interconnected. CDPs integrate first-party data, including customer-supplied data and purchase history, website or app behavior, and marketing response and engagement information, with third-party data on customer interests and shopping behavior, to improve individual targeting.
The brains that drive automated decision making are the advanced analytics models that produce propensity scores for each customer or prospect. These scores define the probability of an individual responding to a specific offer, or engaging with specific content. Whereas standard data models can only pump out messages or offers, modern automated decision-making processes allow two-way communication—collecting and tracking customer reactions and using that information to guide future messaging and offers. More complex decision-making rules, exceptions, or unacceptable variances can be programmed to be escalated to managers (although these exceptions should be less than five percent of all decisions).
The last mile of personalization is content distribution. A good system will use customer and prospect scores to trigger personalized ads and landing pages, and to distribute specific content, offers, or experiences across channels. For example, a telco could personalize the bundle offered to anonymous website visitors based on the type of car they drive, their city, and the stores they frequent. Similarly, an airline can set rules to help automate decisions on the lowest cost offer for a ticket and predict which types of customers will respond to the offer over email, a display ad, or within the mobile app.
For these three “Ds” to operate successfully, companies need to integrate their technology systems, often through APIs, to allow data to flow where it’s needed and decisions to happen in real-time. Doing so allows much of this cycle to learn and adapt in real time, automatically. A series of virtual “pipes” feed response data from customer interactions into the CDP to develop better statistical and event-based models (e.g., response rates based on an event and context). Predictive marketing analytics then make recommendations on which actions drive the highest conversion rates.
Companies that succeed in integrating and automating their customer and marketing data platforms can test, track, refine and optimize themselves in real-time—ensuring that the right offer goes to the right customer or the best leads get routed to the best salespeople. This allows brands to shape their customers’ decision journeys, deepen their relationships, and gain a distinct competitive advantage.
Providing the Right Functional Support
Of course, it’s not enough to gather insights, even at a massive scale. The organization has to be able to act smartly on those insights. Addressing that problem requires changing how things are done, particularly in the following four areas:
- A coordinated strategy. Marketers need to determine where they’re going to play and how they’re going to apply technology to engage consumers. That means defining specific use cases, coming up with schemas, taxonomies, and then orchestrating the right internal and external resources (e.g., agency and technology partners) to manage the process. In many organizations, these responsibilities exist loosely through disparate activities, but they need to be coordinated and managed as a cohesive function to put insight into action at scale.
- Experienced campaign management teams. While automation implies that you can sit back and let the system run itself, there is no autopilot. It takes experienced marketers to set up, deploy, and manage always-on campaigns. Organizations can run systems internally with a dedicated team or enlist their agency to support them—provided they still appoint an experienced individual to liaise with those agencies.
- A sustained commitment to analytics. Data and analytics are the backbone of personalization at scale, but an IT-only project won’t work. Senior leadership commitment and cross-functional involvement are required to support the infrastructure and continuously update the data and advanced analytics models that fuel the decision-making engine. Leaders need to continually test business use cases and adjust the data and tech infrastructure, managing these integrated elements in the way one would a “live entity,” one that is dynamic, fluid and ever-evolving. Because the marketing technology ecosystem evolves quickly, systems need to be flexible so that technologies can be swapped out when necessary.
- A lot of very good content. Content fuels personalization—and someone needs to develop and organize it. You don’t want your “robots” to guide people to stale, irrelevant, or low-quality content. This puts a significant onus on developing a strong content “supply chain” fed by designers, copywriters, animators, and videographers. All content attributes can and should be tested regularly—to refine the look and feel and tone, calls to action, and the value proposition.
Getting to the other side of the “uncanny valley” takes commitment and discipline. But it may be the best option for companies that want to personalize at scale and accelerate their growth in the digital world.
This article originally appeared on the Harvard Business Review site