Part III of IV: The Keys to the Castle: Personalization Rules

March 9, 2015 | John Berndt

I will not show irrelevant content written on blackboardThis blog post is excerpted from a longer chapter in my new full-length book "Personalization Mechanics: Targeted Content for Web Teams of All Sizes," which is now available.

As far as I can tell, when it comes out, it will be the only book on Web personalization written from the perspective of people who actually work on websites (the strategists and teams that build and maintain them), so it’s a much needed resource. I’m excited to be excerpting four sections leading up to the book’s release, to get the conversation started and give a sneak peek into the book’s content. This piece below adapts from a section of the chapter “Better Practices for Personalization Rules.” Check out the full book for much more on personalization rules.

What are Personalization Rules?

Along with appropriate user profiling, personalization rules are really the bedrock of any personalization practice. They can be thought of as contingencies that are set up in a digital experience that dictate what should happen if a certain event or action occurs (i.e., if this happens, then that happens). Both the triggers, and the outcomes, can be quite complex, but they can also be very simple.

Consider a few examples of personalization rules:

  1. If an anonymous user1 connects from a sunny weather ZIP code, show them the “Sun Umbrellas with Puppies” campaign promotion.

  2. Only show this content to a user if they are anonymous and have looked at content that suggests they are already in the “Physics Genius” audience segment OR if they are a known user and the CRM indicates their IQ is known to be over 140.

  3. If a known user has been to the site more than 10 times and completed more than three purchases, show them a completely different layout for the site where you no longer use much of the page to market the brand or position the organization; rather, focus entirely on usability and information seeking behaviors. Shift them forward in the audience segmentation to “regular buyer and site user” and trigger the customer cultivation campaign “Buyer Optimization Four Month Campaign,” which sends them emails.

  4. If a known user is a member of the “Chimney Sweeps Making Under $40,000/year” audience and has seen the “High Volume, Low Cost Dustpan Promotion” on the website more than twice and failed to complete the purchase process, then show them the “Galactic Dustpan Clearance Game Module” and send them an email with a veiled threat about what happens if they DON’T buy. And then run away. Those chimney sweeps are intense.

If the personalization platform is worth looking at, personalization rules are written in a sort of shorthand that is as close to English as possible. The rules are usually assembled with prebuilt conditions, logical operators such as and, or, and not and actions are often brought together in a wizard-style rule builder. This operation is far less technical than is real programming, and any talented marketing or UX professional can be taught to do this.

The basic principles of how rules work and many of the things rules can do tend to be fairly consistent among platforms. How rules are written, how they’re implemented, what they can react to, the peculiarities of their limitations, and what interfaces (sometimes more than one) they are viewed in are, on the other hand, very different between platforms. These differences in turn have great implications for what kinds of scenarios are supported or can be efficiently handled in operations.

Rule concepts are often transferable between these platforms, but platform rule representation, descriptors used, and features supported vary from platform to platform. You must understand the logic behind the rules to smoothly move from one platform to another. For instance, here are three different ways a rule might be stated in different platforms:

Sitecore 8:

Sitecore 8 Rule screenshot

Episerver:

Episerver Rule screenshot

Monetate:

Monetate Rule screenshot

As you can see, they are pretty different, but the idea of stacked conditions, characteristics and outcomes is a commonality. Personalization rules can:

  • Be simple or can compound multiple conditions;

  • Be based on very raw data points about a user’s situation or behavior in the moment or over time;

  • Make use of derivative user categorization using audience segments;

  • Make use of purchase and CRM data; and

  • Trigger a variety of changes in the Web experience, or, when using engagement automation features, send personalized content to users through other channels such as email and SMS.2

equation of banana plus robot equals woolly mammothRule Order Matters

In managing personalization rules, it is crucial to understand how rule and condition sequence affects outcome. Generally, a platform can only present one personalized experience to a user at a time—if multiple rules compete and have met conditions, only one rule can ultimately win.

Consider a page that contains this series of rules3:

— If the user is from Washington, show them item 1; or
— If the user is interested in sports, show them item 2; or
— If the user is on a mobile device, show them item 3; or else
Show the default, item 1.

It is entirely possible that all three conditions will be met. In that case, which item is presented to the user?

Although systems somewhat vary, the answer is generally content item 1. A rule of thumb is that in straightforward cases, the first rule with the first met condition triggers the specified content or experience personalization. Therefore, the order (or priority) in which the rules are stacked directly affects the personalization outcome when the user meets more than one rule condition. This underscores the need to construct well-planned rules.

As you might have guessed, it can be more complicated. Here are three common scenarios that raise the complexity:

1. Complex, Compound Conditions

If multiple conditions are bonded together by logical operators or otherwise related in rules:

(condition 1) and (condition 2 but not condition 3) then content X;
or (condition 4) then content Y; or
else show content Z.

In this example, two conditions are bonded together and must be met before showing content X. (Abstractly and generically, this may sound like programming, I know—which makes sense because personalization rules can be seen as mini-applications…though you don’t have to learn a programming language to create them).

2. Overlapping Conditional Outcomes

In another case, your team might feel it is important that two pieces of content be seen if the user meets two particular important conditions (not one or the other). The easiest way to address this is to personalize two pieces of the page experience with separate rules—assuming there are no conflicts. In more advanced cases, a new rule earlier in the list of rules, one that contains both conditions bound together using the connector “and,” might personalize the experience of the component so that it now shows the correct experience.

3. Rules in Components Embedded in Components with Rules

In most systems, the user experience of a site is to some extent determined by executed programming —integrating HTML user experiences (for instance, “templates,” “views,” etc.) with the content and business logic provided by the underlying platform.4 To most users and marketers, this background programming is invisible, but it raises an interesting consideration: a component on a page being personalized could contain another subcomponent also being personalized.

For platforms (e.g., Sitecore, Monetate, and others) where rules are directly attached to the boxy components of a page, there is a question of which rules fire first. Are they the outer rules of the containing component? Are they the inner rules of the contained, embedded component? Obviously, this can matter for what the user sees, since a rule attached to the outer (containing) component could fire first, replacing the component entirely with something else and leaving the inner component out of luck.

So, the order of rules matters, but so does where they’re attached. In rare cases, it’s also important to understand the underlying modular programming structure of the page, since it influences the order of rule execution.

In general, it is critical during rule development to consider how they will be understood and executed by the platform. To start out, putting the flow down on paper may be a very necessary learning aid to visualize the rule set sequence, and figure out what is going on! Ultimately, the quality of the rules you create and the thought you put into organizing all the pieces will make all the difference between being able to manage complex, effective campaigns that move the needle for your organization, and doing a bunch of stuff that doesn’t quite add up.

That concludes this excerpt from “Personalization Mechanics: Targeted Content for Web Teams of All Sizes." There are many more subjects relevant to rules, including proper rule construction, the issue of abstraction in rules, where different systems like to store their rules (and the implication this has for what the systems can do), and further information on differences between systems. Check it all out in the full book!

Excerpted and adapted from Personalization Mechanics: Targeted Content for Web Teams of All Sizes © 2015 John Berndt.


1The use of this term is colloquial and a bit of a misnomer; in this context it refers to the user who hasn’t converted. Therefore, we aren’t sure who they are in the real world, even if we can gather information about them and profile the hell out of ’em.

2Short message service... or text messaging.

3Or, depending on the system, conditions or compound rules.

4This is true in different ways of both CMS platforms and Optimization platforms.

About the Author

John Berndt

I'm CEO of TBG and I've been thinking about the Web in creative ways since the year it began.

Leave A Reply

comments powered by Disqus