batman figurine near an open laptop with headphones nearby
PHOTO: Victor Xok

Imagine you walk into a big-box retailer. You need some storage solutions to help you organize your garage. You ask the sales associate for recommendations.

“I have just the thing,” he tells you. Then he whips out a Batman action figure.

“That doesn’t make sense,” you say. “I’m looking for storage ideas, not children’s toys.” Yet the clerk reminds you that the last time you were there, you wanted toys. If you come back tomorrow, though, he’ll update your intent in their system and have storage recommendations.

Preposterous, right? Although this wouldn’t happen in-person, scenarios like this happen all the time in the digital world when personalization is based on bad data. In the case above, the data is bad because it’s outdated. Even though it addresses something that may have been true at one point, it no longer is. And when your experience — on a website, in a mobile app, via live chat, etc. — gets “personalized” the next time you visit (rather than in the moment), your intent may well have changed.

In addition to being outdated, data can also be bad — and negatively impact personalization — when it’s:

  • Incorrect: Obviously, personalization based on data that’s misconstrued or entered incorrectly (e.g., into a CRM system) can result in a confusing or even jarring experience.
  • Inadequate: Data is inadequate when it doesn’t tell the whole story. For example, many marketers note they target personalization campaigns based on pages viewed. This is a good start, but an incomplete picture. Was the visitor even interested in what she found? If she’s on a B2B site, for example, at which stage of the journey is she? Which topics and types of content interest her? These other data points make for a more well-rounded, actionable picture.
  • Siloed: If a customer browsed for a watch online, then purchased it from the same retailer in-store, you wouldn’t want to then email him incentives to go buy that watch. When the left hand doesn’t talk to the right, and data gets stuck in silos, customer experience suffers. Effective, 1-to-1 personalization involves bringing together data into a single system for a unified view of each individual across touchpoints.

So, What Makes for Good Data?

On the flip side, data provides a good foundation for personalization when it’s accurate, real-time, in-depth, modeled/analyzed, centralized and actionable. When machine-learning algorithms and predictive models are fed good data, the output likewise is good — that is, a relevant and helpful experience.

But should you wait until all your data is cleaned up before embarking on a personalization initiative? Certainly not. Many companies find that taking a channel-by-channel approach is effective — cleaning up their data and deploying personalization in their highest-priority touchpoint, then moving on to tackle the next.

Related Article: 'Good Enough' Data Will Never Be Good Enough

Data Types for Effective Personalization

Any data that gives a marketer insights about an individual can fuel personalization. Typically, that data falls into one of five categories: attribute data, first-party behavioral data, explicit data, third-party data and predictive scores.

  1. Attribute data: Attribute data describes individual characteristics. They could be digital attributes (e.g., geolocation, referring source, industry, browser/device type, etc.), which are particularly helpful for personalizing experiences for website visitors or app users. Database attributes refer to information pulled from database-driven systems, such as CRMs, email and marketing automation platforms, ecommerce platforms, point of sale systems, and more. You need an identifier (such as an email address, account number, loyalty ID, etc.) to link a person to his data in another system. 
  2. First-party behavioral data: This encompasses an individual’s site-wide and app-wide behavior (e.g., number of site visits or logins, time spent on-site or in-app, time elapsed since last visit, number of purchases made or articles read, etc.) and page behavior (specific pages viewed and number of times). It should also factor in deep behavior and context (active time engaged with particular products, categories, brands, styles, topics, industries, etc.) and campaign engagement (email opens/clicks, push notification dismissals, views of personalized experiences, etc.) for an accurate picture of in-the-moment intent. Finally, transaction data (purchases, returns, registrations, downloads, etc.) is additional, important first-party behavioral information that is generated in digital and physical channels.
  3. Explicit data: You can glean a lot about an individual from your implicit data. But when you’re stumped and need more info, sometimes the best thing to do is ask. Forms and surveys are ideal in these situations. It’s important, though, to deploy surveys strategically. You don’t want to annoy or alienate your audience! So make your forms and surveys brief, and deploy them after you’ve earned the right (e.g., not as soon as someone lands on your site). You should also try to use the information to improve an individual’s experience right then and there, so they see the value. For example, if a website visitor responds to a “what’s your industry?” survey, you could immediately present her with relevant, industry-specific case studies.
  4. Third-party data: Data you purchase from third-party sources, such as demographic information, firmographic data, buying signals (e.g., in the market for a new home) and self-defined attributes (e.g., career data on LinkedIn), can also be used to deliver unique, individually tailored experiences.
  5. Predictive Scores: This kind of data is derived from the other four types, but is important enough to call out separately. Many organizations have their own data science teams generating propensity scores (e.g., likelihood to churn,  purchase or engage in a channel), and some personalization platforms will natively generate propensity scores and affinity scores. 

Related Article: 5 Drivers of Personalized Experiences: A Walk Through the AI Food Chain

The Role of a CDP in Personalization Efforts

The information above represents A LOT of data that can come from A LOT of channels and activities — spanning deep, multichannel behavioral data, as well as data from in-store transaction feeds, loyalty systems, call centers, chat tools, voice of the customer systems, and much more. This is much more — and more complex — data than a CRM can process. That’s where customer data platforms (CDPs) come into play, with the ability to collect, store and synthesize all this data at the individual level, and make it actionable. There are various types or levels of CDPs, including unified personalization and customer data platforms that can apply machine learning and predictive analytics and models to the data to determine and display the best experience to a given individual in real time and across channels.

Related Article: Clearing Up CDP Misconceptions

Garbage In, Garbage Out

Faced with mounds of customer and prospect data, companies today have a duty to use that data responsibly and effectively to generate helpful, relevant experiences. Bad, sloppy data yields bad personalization … which, when you think about it, isn’t really personalization at all. It’s an experience that misses the mark, and can lead to frustration and lost opportunities.

There’s an old saying: “Garbage in, Garbage out.” In this case, if you use faulty or incomplete data for personalization, you’ll get a poor result. So, rather than use garbage, power your programs with good, useful data — for an experience that will make your audience stop and smell the roses.