Todd Fahlberg of Enterprise Knowledge, and Amber Simpson, a Senior Manager at Walmart Academy, presented on November 9, 2022 at the KMWorld Conference in Washington, DC on the topic of Building an Innovative Learning Ecosystem at Scale with Graph Technologies. In this presentation, Todd and Amber share how they’re making it easier for Walmart’s learning organization to manage content used by 2.4 million global associates with a custom Digital Library. The presentation provides insight into the challenges they faced and the lessons they learned along the way, in addition to their approach to design and implement the Digital Library. Todd and Amber also detail how and why they used graph technologies to make certain their solution can continue to scale to meet the needs of Walmart’s massive workforce and evolving business needs.
3. L&D Team Statistics
• +3.5M Assignments
• +4k Video Assets
• +4.3k eLearning Projects
• +5k eLearning Modules
• +6k Question & Answers
• 3 SharePoint Sites with over 700k Files, Records, Assets, Documents, etc.
Walmart Academy – By The Numbers
Traffic
• +80k SharePoint Site Visits Every 90 days (L&D Team)
• +500k Completions per Day (Associates)
4. 4
Our Challenges
Why Is It So Hard to Find Knowledge?
TODAY’S PROBLEMS
• Lack of visibility to current and
future skill gaps within our content.
• Inability to leverage semantics and
search technology for search
optimization.
YESTERDAY’S PROBLEMS
• Lack of knowledge of existing
content.
• Manual processes.
• Siloed learning solutions with
unstructured data, insufficient
search capabilities, and require
institutional knowledge to discern.
TOMORROW’S PROBLEMS
• Need content data model to
support content curation,
recommender systems, and
informational queries.
5. • I’m lost and don’t have time to find it
• When I find content, it's out of date or not for me
• I’m not clear how to help my team prepare for the
future
• I don’t see the value for me or my team
• Learning happens where I work
• Learning suggestions feel personalized & at the right time
• Future skills & learning needs feel accurately predicted
• Learning is intuitively connected to other talent processes
• Data shows the impact of learning in business metrics
Evolving from Reactive & Ad-Hoc to a Connected Learning Experience
Turning this… …into this.
7. Future
Past
• Libraries utilize a card catalog which
contains index cards arranged by main
classes covering the world’s knowledge.
• Every artifact has a specific index card
with standard attributes that includes a
bibliography, subject, call number and
citing/references.
• Users navigate through the card catalog
to access or “call” a specific artifact.
• With semantic search, users can
navigate the digital library to quickly
find, discovery, view, access, audit,
curate, etc.
• Every learning artifact has a specific
record with standard metadata that is
auto-tagged and systematically related.
• Data about our learning artifacts are
arranged by main classes covering our
learning portfolio.
What Did We Want?
“The sole purpose of a library is to provide resources and services to meet the needs of individuals
and groups for education, information, personal development.”
(IFLA/UNESCO Public Library Manifesto, 1994)
8. The Knowledge Graph Business Case
Our problems required a technology
that can:
1. Meet current needs and
continue to scale as Walmart
grows.
2. Manage highly interrelated data.
3. Help to organize sparse, siloed
data.
4. Remain flexible as new
integrations and data models are
introduced.
5. Keep up with evolving data and
content.
Why Graph Database Over Relational?
Relational Database
A data model featuring rows,
attributes, and constraints.
Strengths
• Great for tabular data with a
consistent structure, a fixed
schema, a simple relationships
(one-to-one/many).
• Most common data model in use
for databases across all major
industries.
Weaknesses
• Not able to easily represent
complex relationships between
multiple entities (relationships are
not “objects”).
Strengths
• Purpose built to manage
relationships (relationships are
“objects”).
• Model allows flexibility in schema
design with low disruption to rest
of database for adding new data
or relationships.
Weaknesses
• Not meant to bulk read queries
that request large amounts of
data across the database
(decreased performance).
Graph Database
A data model featuring entities
and their relationships.
10. • Google “like” Search
• Faceting by ”Content Types”
• Powered by Search
Optimization Engine
Our Solution – L&D Team Perspective
One Centralized Digital Library that is the
Source of TRUTH for Walmart’s Learning Portfolio
11. Our Solution – L&D Team Perspective
Faceting
Leveraging
our
”Learning
Taxonomy”
Quick
Access to a
Content
Details Page
Identifies
“Related”
Learning
within Search
Results
Connects to
“Source”
System for
Quick View
One Centralized Digital Library that is the
Source of TRUTH for Walmart’s Learning Portfolio
12. Our Solution – Architecture Perspective
Learning Developers
Semantic Search
Taxonomy Management
Knowledge Graph
1 – Learning & HR
Repositories
• Learning Content Database (LCD)
is the source system that is an
index for all learning content and
its metadata.
• Functions as a service to benefit
other Walmart systems and
initiatives.
2 – Semantic Search
• Solr, powered by the LCD’s
Knowledge Graph and
taxonomies, provides Academy
and its internal and external
facing systems a Google-like
search experience.
3 – Knowledge Graph
• Manages complex, semantic relationships
between learning content and systems.
• Enables Academy to personalize the
associate learning experience.
Learning
Content
Database
CMSs &
DAMs
LMSs &
Authoring
Credentials,
Badges, Skills,
HR Core
4 – Associate Learning Experience
• Learners experience a Google-like search
and receive course recommendations based
off their role, learning history, and earned
badges.
2 3
Learning Experience Solutions
Academy App ULearn
Search
4
Our Associates
1
13. KM World
Explored the domains
of Taxonomies and
Semantic Technologies
Knowledge Management
& Taxonomy Design
• Standards & Governance
• KM Strategy & Ways of Working
• Lots of Research & Value
• Strategy Development
• Outlining Roadmap
• Change Management
• Teaching Moments
Launched MVP
Continue to Iterate & Integrate
SINCE MVP
Integrations with Multiple
Platforms
• LMS & WCMS Platforms
• Taxonomy Management & Graph
Database
• Search Engine Optimization
(SEO) Tool
Implemented
• KM into our existing WoW
• Governance & standards
• Learning specific taxonomies
Upcoming!!!
• More Integrations, More
Taxonomies, More Content Types
ATD Tech
The Journey Thus Far…
Benchmarked with
Learning Professionals
Q4 2019 …Q3 2020 Q1 2021
Q1 2020
15. 1. Identify your organization’s standard processes for building content.
v Archiving and storing content artifacts?
v Naming (unique ID) content artifacts?
2. Identify your organization’s current learning platforms and define the standards &
build a governance.
v LMS, CMS, WCMS, DAMS, HR, etc.
3. Identify & define how your organization describes learning content? This is your
“seeded taxonomy.”
4. Implement standardization, at the source, for adding the “seeded taxonomy” to the
content.
v Learning Objective, Level, Audience, Duration, Skill, Competency, etc.
Getting Started
16. 1. Show Yourself GRACE!
v Accept that perfection doesn’t exist, but continue to strive for it…
2. Taxonomies, Ontologies, Oh My!
v Keep it simple. This shouldn’t be complicated.
3. Meet people where they are!
4. Training, Training, Training, Training.
5. Clean Data is a must but waiting on perfection will delay your start.
6. Find a place that you can impact. Identify the problem and jump in.
7. Identify a trusted partner(s) who is willing to teach, mentor, and provide guidance.
8. Stakeholder, Business, and Tech alignment.
Lessons Learned