It's the smackdown of the season, a battle like no other, and it's happening in the boardrooms, corner offices, and brainstorming sessions of businesses large and small. When do you automate and when are you, the human marketer, an appropriate relevance-bestower, able to successfully implement and fuel a meaningful personalization campaign?
For today's digital business striving to deliver personalized experiences-that just-right content or product to the right person at the right time-tapping into an automated decisioning and personalization system is likely the key to achieving the type of scale at which successful commercial Web and mobile sites and apps operate.
A few months ago I had the privilege of presenting alongside Dr. Keith Devlin, co-founder and Executive Director of Stanford University's H-STAR Institute and Media X Research Network. He's also "The Math Guy" on NPR and an accomplished author and speaker. During our presentation, Dr. Devlin challenged the audience-all well-regarded, highly trained digital marketers-to think about whether they were capable of successfully predicting what their customers want. For some in the audience, this could literally be millions upon millions of unique consumers. He then took it a step further-why even consider a question like that? It's borderline arrogant, Dr. Devlin proclaimed, to think that human marketers can comprehend those kinds of numbers, let alone understand the motivations and needs of that many individuals. You're good, but you're not that good. That energy, that brain power, and that know-how can be better channeled elsewhere, in areas of the personalization process that will actually move the needle.
I tend to agree with Dr. Devlin. I don't believe that the human marketer can go it alone, not when relevance at scale is concerned. Do you truly have the capacity to be spot-on with your messages, offers, and product recommendations? Maybe sometimes . . . but all the time? Without missing a beat? Being relevant up to and including the customer's last millisecond interaction is critical to success and, realistically, without leveraging at least some level of machine-driven personalization, customer experiences will no doubt be compromised, at least to some extent. And when consumers don't feel the personalization warm and fuzzies they crave, they'll quickly find a site that can deliver.
So are today's digital marketers ready for automated personalization? The answer is somewhat mixed, reveals Adobe's 2014 Digital Marketing Optimization Survey. But my answer is a resounding yes-yes digital marketers need to get on board, and yes they need to get there-fast. Although few organizations have fully embraced automated approaches to personalization, most want to learn and expand. When they reflect on their own optimization maturity, they tell us this over and over. They don't want to overstate their capabilities, but they have a desire to experiment, experience, and just do.
Feeling anxious? No need, but just a few things to know
When it comes to understanding approaches to personalization, it's important to wrap your head around the vernacular. Don't worry, I'm not about to suggest you retake that mandatory college statistics course. Think of it more as a primer-one that will help you when the conversation veers from topics that are in your comfort zone (branding, demand generation, campaigns) and into the realm of data science. Of course, as I've discussed before, personalization is not-and cannot-be just about data and technology. Even automated personalization needs a human touch. Watch for a post in the very near future on how and when to get started, sprinkled with a few best practices. For now let me leave you with this quick and dirty glossary that will help you sound like a personalization pro:
- Affinity engine: Pattern-matching technology that predicts affinities, such as likelihood to prefer one product category or brand over others based on implicit and explicit visitor actions and behaviors.
- Algorithm: A computed logic process to produce a desired outcome. We're talking about the math that underlies most of what we experience in computing, from pretty simple computations to rather complex ones that would take a human significant time.
- Behavioral targeting (BT): Either automated or rules-based, BT involves personalization of content, or otherwise altering digital experience, based on any visitor variable or combination of variables with the goals of eliciting a desired response.
- Click-stream data: A data record of a user's sequential Web activity. This includes webpages visited, how long he or she stays on page, and the sequence of page interactions and paths the user follows.
- Collaborative filtering (aka "wisdom of the crowd"): A method for predicting the interests of a visitor by collecting preference information from many visitors. If person A has the same interest in an item as a person B, A is more likely to have B's interest on a different item.
- Cross-selling: Presenting site visitors new or complementary content/products/services based on consumption behavior and intent.
- Decision engine: In digital marketing this involves use of advanced algorithms/machine learning to predict content or offers based on analysis of a wide range of visitor interaction. Ideal for locations on digital properties that have large traffic volumes of diverse, harder-to-predict visitor types.
- Ensemble learning: A machine-learning technique that combines several algorithmic approaches or models (an ensemble) and makes them compete to determine the best approach.
- Look-alike modeling: A great way to put Big Data to use, look-alike modeling pulls data from first-party info-think purchase history or expressed preferences-to create consumer profiles that can be used to market to other users, specifically anonymous or otherwise unknown users.
- Machine learning: Machine-learning systems learn from training data sets and are used to make decisions and predictions, typically in environments with very large volumes of data such as large commercial websites.
- Progressive personalization: The process of building and leveraging a progressive profile to deliver a personalized customer experience from one touchpoint to the next. For example, today's (and definitely tomorrow's) customers want a persistent experience as they move from mobile phone to physical store to tablet to interactive voice response (IVR) to kiosk-and back again!
- Recommendation system: AKA recommenders, these systems can range from simple collaborative filtering to more complex machine-learning approaches designed to make relevant suggestions to visitors/app users. Ideal for product cross-selling and article suggestions to optimize engagement.
- Searchandising: Merchandised search results, landing pages, banners, and cross-sells based on business rules that augment or override natural relevance ranking, and reflect merchant/marketing priorities and shopper analytics.
- Segmentation: Creating buckets of customers or visitors based on one or more similar characteristics (demographics, device type, geo-location, browsed categories, etc.). In digital marketing this is usually rules based and influenced by marketing strategy, personas, top revenue performers, and analytics.
- Targeted discovery: Personalized site search and browsing experiences based on targeting rules and visitor profile attributes.
- User modeling: A mathematical practice used to make predictions about a site visitor's behavior.
- Visitor profile: A detailed yet anonymous representation of a site visitor created by collecting explicit and implicit information about the visitor through the individual's digital interactions.
- 360 personalization: Leveraging multichannel and/or third-party data for profile augmentation, 360 personalization reflects a customer's complete digital fingerprint.