By using a modular content platform, marketers and content creators gain the ability to tailor information and messages that adapt to what users will be interested in. This content tailoring is commonly known as personalization.

Personalization concerns deciding what content to show to whom under what circumstances. Despite being widely discussed, personalization is not understood by everyone in the same way. 

Most people agree that personalization is highly desirable for both enterprises and their customers, provided it is accurate and not clumsy. 

At the same time, many digital professionals will also admit that personalization is challenging. There are plenty of failed personalization projects that didn’t change business outcomes or which produced distracting or annoying experiences for customers.

Yet fewer folks agree on how personalization is supposed to work and what it is aiming to achieve. How does content adapt to different individuals, and what content needs to adapt? Personalization can seem like a black box. 

Personalization is not a magic technology. It requires both thoughtful planning and an accurate understanding of users.

Some people wrongly see personalization as simply prospecting for customers, as when the goal is to find out who would be interested in clicking on an ad. That isn’t personalization because the content is invariant. Personalized content, by definition, involves varying the content in some way to meet the needs or interests of different kinds of customers.

Personalization, thus, is the intersection between the content and the content user. It can be approached from either side. Given a certain individual, what content will appeal to them? Or alternatively: given certain content, what kinds of individuals will find that appealing, and are they who you need to reach? In the latter case, the content could change to attract a different or broader range of individuals. Personalization involves learning, often with the aid of data and machines.

Personalization is not a single approach to deciding the right content for an individual. Rather, it relies on a range of orchestration methods. Broadly speaking, two core approaches to personalization are prevalent, which have different intents.

The first orchestration approach is what we’ll call “matching.” The second can be thought of as “directing.” Both approaches take advantage of modular content.

Personalization through matching

Matching provides a personalized mix of content items that will express the “best match” for the user, considering both the types of items they need and the version of the items that are most relevant to them.

The mix presented will reflect a prediction about what will interest the user, which could be based on either the content creator’s understanding of a user segment and context or data about user preferences developed from multivariant testing.

Customer parameters	Content variables
Customer characteristics Customer context Explicit customer intent, such as search queries	Topic Message theme Geographic coverage Journey relevance Content format Channel

Matching typically involves pairing a customer segment that describes an individual (defined by their characteristics and context) with the most relevant items based on the content’s topic, geographic coverage, journey relevance, content format, or channel.

The matching is a form of curation. It implements rules for what to present to a given customer in a predefined circumstance. A common example is when a customer is searching for content about products with certain attributes and is directed to a special landing page featuring curated products that match those search terms. In other cases, the customer’s intent is inferred by what’s known about their context and circumstances, such as the source or path that prompts the customer’s visit to the curated content.

Matching orchestration is common in scenarios where the publisher is presenting many options for further exploration or is bundling benefits associated with a single call-to-action. Landing pages rely on matching, as do curated home pages.

The match might be based on many customer parameters but just a few content variations when there’s overlap in what customers want. Alternatively, a few customer parameters could generate many permutations for relevant content if there’s little overlap in what various customers need.

When the number of expected scenarios is limited, it’s often possible to create in advance the required variations as larger curated content items. When the number of combinations gets a bit bigger, variations can be generated dynamically in real-time using the code in the UI frontend framework by querying the content’s modular elements and taxonomy tags.

As the number of customer parameters and content variables grows, publishers can use recommendation engines that rely on AI to make the best match.

Personalization through directing

Directing provides a way to adapt content when the customer’s intent is unclear. Rather than choosing a fixed set of items to match the known preferences of customers, the directing orchestration is more focused on discovering those preferences. It qualifies the readiness of customers to take action associated with the content presented.

The ideal mix of information and messages to communicate to an individual often depends on multiple factors that can’t be predicted ahead of time. Their reaction to an initial message or option will determine what’s the most relevant follow up content to provide them.

Directing personalization tries to learn from a customer’s behavior. Enterprises provide information and messages based on the customer’s prior expression of interest. They’ll present one set of information and messages if the customer indicates interest through their behavior or choices, otherwise they’ll present an alternative to see if that elicits an action.

With directing orchestration, content decisions are related to customer decisions. Is a change in content likely to change customer behavior? And if customers are deciding incrementally, how does the content need to adapt?

The table below illustrates examples of how customer parameters, content variables, and customer behavior can interact. In the initial encounter, the content adapts to some parameter related to the customer. The orchestration then evaluates how customers interact with the presented content to provide follow up content based on those signals.

Customer parameters	Content variables – type and variation (Round 1, the invitation)	Customer behavior	Content variables (Round 2, the follow up)	Customer decision
Referral source Journey path Customer status	Promo Recommendation Offer Teaser	Clicks to learn more Watches video Opts in with preferences	Elaboration Alternative promo Offer to remind later	Acts Defers Dismisses

Directing orchestration can be used for both marketing and customer support scenarios. For example, an appeal to customers to upgrade their loyalty app could be presented as different content types: as a promo about new features, a teaser about missing out, an offer promising extra benefits, or a warning about losing access, depending on how you want to frame the issue. And each type could be written in different ways. The content alternatives could be relevant to customers in different situations and would likely prompt distinct customer behaviors.

Directing orchestration is common when the publisher needs to present a complex message that involves qualifying interest or eligibility. It’s used when presenting content sequentially involving questions, options, or offers, such as purchase eligibility scenarios. It’s also useful when wanting to separate content centered on promoting attention and attraction from the associated content addressing the details for customers to consider. 

Compared to matching orchestration, directing orchestration addresses more uncertain situations. While the publisher is likely to know something about the customer’s interests based on their prior relationship or behavior, they will be less certain of exactly what the customer will be interested in when they encounter a message. Customers may choose to explore more or not. After encountering additional information, they act right away, decline to act, defer action, or choose an alternative action. Each of these decisions will prompt a separate set of associated messages and information to present.

The goal of directing orchestration is to apply a set of rules to configure information and messages that generate the most interest with customers. It involves learning:

• Which kinds of people are most interested or receptive to specific appeals or requests?
• What combinations of information and messages resonate?
• When should messages be presented (timing, sequencing, journey stage)?

Because the content directing involves multiple variables associated with messages, information, and their sequencing, this kind of personalization is open to wide experimentation.

Comparing orchestration approaches in personalization

If we compare matching and directing orchestration, we can see they address different scenarios in terms of what’s known about the needs and interests of individuals and how decisions can be made concerning the content to present to them.

	Matching orchestration	Directing orchestration
Knowledge of customer	Higher certainty of customer preferences	More uncertainty about customer preferences
Decision inputs	Expressed or presumed intent	Behavior signals
Decision parameters	Categorization (segment, referral source, scores)	Conditional rules
Decision basis	Prospective	Adjusts based on user feedback
Personalization output	Single bundle mix	Branching, multi-stage content interaction

While matching and directing have different emphases, these approaches are not mutually exclusive. Some personalization tools will blend these approaches, and many enterprises adopt a hybrid approach to personalization, depending on the content type and tasks associated with the customer journey.

By adopting a composable approach to AI, enterprises can select the personalization tools that will best support their goals for specific content.