What drives changes in data models, and how can we best manage those drivers and/or changes?

June 24, 2020, 3pm Eastern

Email eva@makingbetter.us  to be added to the invite 

Team

• Host: Aaron Silvers

• Community Manager/Secretary: Eva Rendle

• Producer: Megan Bowe

• Attendees:

  • Jonathan Poltrack (Veracity)

  • Tom Creighton (Veracity)

  • Kristi Eager (Boeing)

  • Brian Duck, DuckWorks LLC

  • Jeff Clem

  • Lang Holloman

  • Marcus Birtwhistle (ADL)

  • Ivan Martinez-Ortiz

  • Tobi Echevarria (Veracity)

  • Jason Haag (Veracity)

  • Viktor Haag (D2L)

  • Christina (on phone)

  • Rob Chadwick (Veracity)

Please join us for a discussion on the topic, “What drives changes in data models, and how can we best manage those drivers and/or changes?”’

Slides: https://1drv.ms/p/s!Av4nfZFOBTgagVGDm5LgvQ4SuYcC

Discussion:

•Objective:

Through regular, open discussions, participants will help identify enabling and/or disabling policies and practices, as well as technical considerations, in leveraging xAPI Profiles (from one or a network of xAPI Profile Servers).

•Exit Criteria:

TAGxAPI will submit an IEEE LTSC technical report for publication, documenting learnings from this series of discussions in Q3 2020

Data Models:

• Account for the data to be generated and used by a system or systems, both logical and physical.

• Provide names, definitions and notes describing the data model elements.

• Accompany implementation guidance, such as data types, keys, indexes and views that are critical to the challenges addressed by the data model(s).

• Describe business rules in terms of relationships (referential constraints) and values (restrictions

• Identify factors to secure data, data sources and any rules to be maintained.

• Define governance plans, including enterprise-level collection details (Master Data Management), data quality and retention policies.

Benefits of Data Models

•Design

•Manage Redundancy

•Integrate & rationalize

•Increase quality

•Implement & Maintain

•Increase discoverability

•Improve comprehension

•Data dictionaries

•Business glossaries

Business Glossaries

• Maximize understanding of the core business concepts and terminology of the organization

• Minimize misuse of data due to inaccurate understanding of business concepts and terms

• Improve alignment of the business organization with the technology assets and technology organization

  • Help provide a source of truth for stakeholders – what they are looking for and why

• Maximize the accuracy of the results to searches for business concepts and associated knowledge

Drivers that Change Data Models

• Deleting something

• Adding something

• Renaming something

• Changing the data type of something

• Changing the constraints or rules for something

Responsible Parties

Different organizations approach data models differently, so drivers for data models, and changes to existing data models, vary.

• Who understands the value of the data models-

  • Use of data models across projects

  • Strong variance between organizations and who in the organization cares about them

Communication Gaps

• Only 45% of projects tend to include a data modeling team early enough in the processes to influence development

• On 3% of projects surveyed, the data modeling team was included only after decisions were made that could not be changed

• 28% of organizations only document a data model AFTER delivery of an application

• In 25% of projects where a data modeling team is available, the team finds out about the project too late.

Shared understanding required – gaps become exacerbated with lack of communication

Data Model Change considerations

• Must work in concert with any workflow – not just waterfall; not just agile

• Check-in and Check-out capability

• Associate data model changes to requirements such that they can be controlled for

• Audit trail of changes: what was done; why; compliance with governance policies/rules

• Ability to compare models to databases and other models and identify changes that need to be merged into the source of the target

• Capability to create branches from a model baseline and merge them back in or roll back to restore a previous version

• Ability to generate the necessary code to implement the desired changes

How do xAPI Profiles Fare?

• Supports different working styles and workflow approaches to data modeling

• Enables machine-readable data dictionaries via the JSON-LD representation of the profile.

• A proper xAPI Profile document (human-readable) serves as a business glossary

• When managed  through an xAPI Profile Server…

  • We can track *why* changes are made, in addition to *what* changes are made

  • We can correlate changes with development changes

  • We can implement model version control

  • We can provide an audit trail for compliance or governance

Discussion question-

What are some challenges to consider when establishing or changing a data model?

Jono Poltrack:  SCORM example – SCORM has a data term of SCO (when written, SCO was created)

• Data perspective – everything is a SCO

  • Chapter, learning objective, lessons, modules – whatever you call it is fine in your organization

 

Realizing that Activity type (lesson) is overloaded in xAPI

 

Abstract concept (lesson – certain set of data)

         Technical definitions vs instructional designer is trying to create (lay vocabulary)

  Technical artifacts are created

 

Aaron Silvers: Instructional design does not have same set of vocabulary across the industry, little bit like the wild west- no normalized training, people come into it from different ways

 

Need common vocab or labels

 

Jono Poltrack: Elearning example:

• Lesson – technical perspective – needs particular data around it (score, pass/fail, completed)

• Needs rules around it – how underlying data should look, but not putting constraints on organizations and their language

 

Kristi Eager:  Video Profile example

• New products come out and they reference old versions of profiles

• How likely will data model need to change, capture data differently over time?

• If something changes, how make sure some data point doesn’t get abandoned

 

Jono – IEEE could be helpful in this problem, especially with the more general profiles like audio, video

         Create a standards process

         9272.3 – standards for Profiles

 

Profile server administration –

• Are profiles verified - 

  • Yes, has verified Profile – like a verified twitter account

• Verified over time

 

Kristi: How long is xAPI data valuable for?

• Landmark statements?

• Currently all statements are not deletable

• As more hardening enterprise management and analysis comes into play – probably a time limit, or maybe a state or maturity.. longer term question