Content Personalization in Drupal with Drupal WEM Project

CONTENT PERSONALIZATION
Drupal Camp MA

January 19, 2013

Jason Yarrington | www.digitalbungalow.com | http://www.linkedin.com/in/jasonyarrington | @jasonyarrington | jyarrington@digitalbungalow.com

INTRODUCTION
Drupal Camp MA


WHO AM I

Jason Yarrington

Vice President, Professional Services
Digital Bungalow
» I have been working on web based marketing solutions throughout my career.
» Psychology – Sociology – Statistician with the Census Bureau
» Did my time at web startups in Boston
› Worldstreet
› Smartbargains
› LifeOptions
» At DB we focus on building engaging marketing web sites.


CONTENT PERSONALIZATION

» Define the problem – making content more relevant to individuals
» Have a goal
» How do we implement this with Drupal


THE PROBLEM
Drupal Camp MA


ENGAGING CONTENT THE PROBLEM 6

» When looking at site traffic, most analytics are done in aggregate.
The aggregate results show us how the ―Average User‖ engages
with the site.
» However, understanding analytics and content needs in aggregate
is no more useful then utilizing the average temperature for a year
in deciding what clothes to wear on a given day.
» As a result, most websites don’t feel very smart to the user.


CONTENT PERSONALIZATION 7

The Wor


ENGAGING CONTENT TARGETTED CONTENT 8

WE DESIGN FOR PERSONAS
» The products we sell are designed for specific people
› Clothes for men, women, infants, children
› Cars for environmentally conscious, trucks for middle age tech
executives that wish they spent more time outdoors
» Content is designed for specific people
› Blog articles based on interest
› Technical articles for the computer programmer
› Business articles for the project manager
› Design articles for the UX person


WEB ENGAGEMENT MANAGEMENT NORDSTROM 9

Example:
Nordstrom.com
Homepage
Not Smart!

Use Case:
Semi-Frequent
Male Shopper
(cookied)

Smart!


WEB ENGAGEMENT MANAGEMENT NORDSTROM 10

Example:
Nordstrom
Email
Marketing

Wasted
Use Case: Impression
Semi-Frequent =
Male Shopper Not Smart!


DATA DRIVES ANYTHING AMAZON 11


NETFLIX REAL MONEY IN THE GAME 12

NETFLIX CINEMATCH
» In 2009 Netflix ran a contest to see if developers could create algorithms that were
better than there’s at predicting someone’s ratings based on a training data set of
ratings.
» The goal is to see if an algorithm can better predict whether someone will like a movie
based on previous ratings.
» If someone could make a 10% improvement, there was a prize


PERSONALIZED RECOMMENDATIONS NETFLIX 13


GETTING UNREAL 14

When we put a site together, even if we narrow in on a segment or
persona, there is still a lot of variety in interests for people of those ages.

But the good news is, they want to tell us their interests


WEB ENGAGEMENT MANAGEMENT 15

The promise
» If we present users with content that fits their preferences will
they be more likely to return?
» If the visual elements of the site respond to the user will they
relate to the site and content more?


MARKETING 101
Drupal Camp MA


MARKETING BACKGROUND 17

A little marketing 101 the way we think about it.
» AWARENESS: How do I make you aware of my product
or service.

» ENGAGEMENT: Now that you know about my product
and service, how do I get you to learn more.

» CONSIDERATION: You know a lot more now, how do I
get you nudged towards making a decision to purchase.

» PREFERENCE: This is very close to consideration.
But, it’s a preference for my brand over another. Do you
want my product or service rather than my competitor.

» CONVERSION: Time to purchase.

» LOYALTY: Come back and purchase again.

» ADVOCACY: Recommend our product or service to
another person



Can we make a site more engaging?
or service.






another person


CASE STUDY 19

Realforme.com
» Popular health and wellness site with content from over 20 bloggers.
» But everyone sees the same content, regardless of who they are.
» Login is just a sort of pay wall. There is no advantage to me to joining this site.


Humana Program Background

Realforme.com was a
successful acquisition
program.
Over 370,000 members had signed up
through a very successful multi-channel
online and offline campaign over three
years.
• Members of the site were much more
likely to become—and stay—Humana
subscribers.
• The program was awarded Best Web-
Based Customer Retention and
Loyalty Campaign by the CMO Council

But, we don’t see people come back often.


ENGAGEMENT STRATEGY

DEFAULT VIEW

Segment Attributes:
• Brand new visitors to the site with no click activity
• Users who have yet to be modeled
WEM Output: Default version of the page


ENGAGEMENT STRATEGY

HEALTH SEGMENT

Segment Attributes
• Demographics: Male or Female, Under 60
• Tracked Site Activity: Users who have predominantly clicked on health related content.
WEM Output: Web and email featured content is weighted toward new health content.


ENGAGEMENT

RETIREE SEGMENT

Segment Attributes
• Demographics: Male or Female, Age 60+
• Tracked Site Activity: Users who have predominantly clicked on retirement related content.
WEM Output: Web and email featured content is weighted toward new retirement, leisure, & finance content.


CASE STUDY

FAMILY SEGMENT

Segment Attributes
• Demographics: Female, Age 30-60
• Tracked Site Activity: Users who have predominantly clicked on Family & Social content.
WEM Output: Web and email featured content is weighted toward new Family & Social content.


ENGAGEMENT

RESULTS

• WEM has been a huge success. User engagement is up across:
 Total visits: 36%

 Unique visitors: 46%

 Page views: 72%

 Visit duration: 74%

• Our client is thrilled!
• This project is a shining example of our ability to innovate beyond
what we thought was possible, and to craft new technology that
delivers on our vision.


ANOTHER EXAMPLE
Drupal Camp MA



Can we make a site more engaging?
or service.






another person


BREAST TOMO TARGET MARKETS 29

In order to convey a personalized experience for each user group on the
website, we will create strong calls to action and highlight relevant content for
the primary target markets.


BREAST TOMO WHAT IS WEM? 33

» Medical device sales require the development of highly targeted
quality content to help purchase influencers and decision makers
alike understand the benefits of the product.

» Web Engagement Management (WEM) is the practice of
customizing and personalizing website content for users to make
the website experience more relevant and compelling. This
encourages users to spend more time on the site and to be
presented with more relevant content on each visit.

» In order to do this, we determine visitor segments and set up visitor
scoring modules, dynamic content, and analytics, in accordance
with the user and business goals for the site.


APPROACH
Drupal Camp MA


BREAST TOMO WEM STRATEGY 35

Users to the site will be analyzed on two axis:

Engagement Level User Type

How engaged a user is with the site Based on what content the user
and brand measured by the quantity interacts with, we will categorize the
and quality of their interactions user into the target segments


TRACKING AND SCORING 36

User engagement will be a measure of all the user’s prior activities on the site.
» User engagement will be tracked via
pre-determined site actions. Engagement Scoring
» The content for these actions will have
established values which will reflect the
weight each action has on the user’s
overall engagement level.
» Upon completion of any one action, the
point value will be added to the user’s
overall engagement score.

Action Content/Asset Point Value
View… Page 1
Video 20
Download… PDF 50 Conversion
Click… Promo Tile 25


ENGAGEMENT-BASED USER PATH 37

This is Tim.
When Tim…

…visits the website, we give him a point.

…downloads a white paper, we give him 10 points.

…watches a video, we give him 5 points.

…and reads articles, we give Tim even more points.

Methodology: Once Tim reaches a certain engagement score, we move him into the next
engagement tier or segment.

Result: Tim will now receive content specified for his particular engagement level.


PERSONALIZATION USER TYPE 38

Role categorization is based on the type of content the user interacts with most.

» Individual content pieces will be tagged with one or Radiologist
more role types.
» When a user accesses a piece of site content, the
user will be given a point towards that particular role.
» Upon reaching an initial threshold of at least one
standard deviation from another role, they will be
assigned a role.
» The user will then be served content developed C-Suite
specifically for their user type.


BREAST TOMO ROLE-BASED USER PATH 39

This is Craig.
When Craig…

…reads an article tagged for radiologists

…downloads a white paper tagged for radiologists

…watches a video tagged for radiologists

…lands on a page tagged for radiologists, we give Craig a point towards
that particular user role.

Methodology: Once Craig receives 15 initial role points, we will assume, with a high degree of
confidence, that Craig is a radiologist.

Result: Craig will now be targeted with specific content for that role.


BREAST TOMO WEM CONTENT PLACEMENT
The following real estate will be utilized on the homepage to personalize the
experience for the user’s role and engagement level.

Role-Based Content

Role-Based Content

Engagement-Based Content


THE CODE
Drupal Camp MA


Web Engagement Module

http://drupal.org/project/wem

Initially we did this all with
custom code.

We took the practices we did in
this project and abstracted them
into the WEM module.


PROJECT/WEM 43

KEY FEATURES

» This module allows you to track user events as they occur throughout the site. By
doing this, you are able to customize a user's experience to be unique.

» The Engagement module can utilize the Google Analytics module to help track
events. Any event that you track in this module can be sent to Google Analytics.


ENGAGEMENT API 44

ENGAGEMENT API

» The Engagement module tracks a user's page views and profile updates by default. You can create custom events very
easily in PHP and JavaScript.

» Track custom events - In PHP:

engagement_event_track($event_name, $event_value = NULL, $event_data = NULL, $uid = FALSE);

» In JavaScript:

Drupal.engagement.track(event_type, event_value, event_data);

» In HTML:

You can track events on any HTML element by using the data-engagement attribute.
To streamline the process, you must use the engagement_data_attribute($event, $value, $data, $bypassQueue) function.
<a href='#' <?php echo engagement_data_attribute(array('click'), 'submit_form', 'registration_form'); ?>></a>

This 'a' tag will automatically be tracked on it's click event with the inserted event attributes.


WEM 45

EXAMPLES
Let's say we want to change a user's site experience based on the articles they read.
We could make a tag pool of the articles that they view.

First, let's set custom tracking based on tags. In your hook_page_build():
// Get the tags for this article.
$n = menu_get_object();
$tags = $n->field_tags[$n->language];
$tag_array = array();
foreach($tags as $tag){
$tag_array[] = $tag['tid'];
}
engagement_event_track('article_view', current_path(), array('
tags'=>$tag_array), $user->uid);


WEM 46

QUERIES
The Engagement module comes with easy-to-use functions for doing common querys.
Doing a custom query is easy too because all data is stored locally.

$events =
engagement_get_events('article_view', NULL, NULL, $user->uid);
$tags = 0;
foreach($events as $event){
// Count tags and put in array.
foreach($event->data['tags'] as $tag){
$tags[$tag]++;
}
}

We now have a user's total engagement for all tags in the $tags array.
We can customize content for the user with this information.


WEM 47

PERSONALIZING CONTENT
We now have a user's total engagement for all tags in the $tags array.
We can customize content for the user with this information.

if($tags['tech'] > 100){
// Give user the 'tech' badge or something..
}

In JavaScript, we could track when a user hovers over a certain link:
Drupal.engagement.track('mouse_hover', 'secret_button', hover_time);

If you have the Google Analytics module enabled, these events are automagically sent as
custom Google events.


WEM 48

TRACK NODES


WEM 49

TRACKED USER ROLES


WEM 50

FILTER WHAT PAGES TO TRACK


WEM 51

REPORT ON TRACKED EVENTS


WEM 52

CREATE SEGMENTS


WEM 53

CREATE SCORING RULES ON EVENTS


WEM 54

COMPARISON SEGMENTS


WEM 55

CREATING SEGMENTS


AUTOMATING THIS

RECOMMENDER API

Recommender API implements various
algorithms to build recommender
systems. It provides easy-to-
understand, easy-to-use, fully-
documented APIs for other Drupal
content recommendation modules (See
a list below). It also provides a unified
approach to configure & execute the
recommender algorithms, and to display
results to end users.


AUTOMATING THIS 57

Recommender API can take your viewers behavior to recommend
additional content
» Browsing History Recommender: "Users who browsed this node also browsed";
personalized node recommendation based on users' browsing history.
» Commerce Recommender: Personalized products recommendations based on
customers purchasing history.
» OG Similar groups module: "Users who joined this group also joined"
» Ubercart Products Recommender: "Customers who ordered this also ordered" and
personalized recommendation "Recommended for you"
» Fivestar Recommender: User-to-user and Item-to-item (used in Amazon.com)
recommendations for Fivestar module.
» Points Voting Recommender: User-to-user and Item-to-item (used in Amazon.com)
recommendations for Up/Down and Plus 1 modules.
» Flag Recommender: Recommender for the Flag module.


58

Jason Yarrington
THANK YOU @jasonyarrington
jyarrington@digitalbungalow.com


Content Personalization in Drupal with Drupal WEM Project

More Related Content

Similar to Content Personalization in Drupal with Drupal WEM Project (20)

Content Personalization in Drupal with Drupal WEM Project

Editor's Notes