Skip to content

Safely Observing Human Systems

RFC 2119

OpenO11y has very strong opinions on how to observe human systems ethically. While acknowledging it is social advice, this guidance is considered a specification. Thus, we use certain keywords in accordance with RFC 2119.

Guiding Principles

Many of us typically think of Observability in the context of observing a system of software (e.g. a collection of microservices that form a single product). Some of the advice provided herein describes how to observe human systems like teams or development organizations. This is a much more sensitive activity. OpenO11y offers these guiding principles to observing human systems ethically.

  1. When observing human systems, individuals MUST feel and be safe while being observed.
  2. Telemetry data MUST be anonymized and/or aggregated to protect the privacy of the individuals.
  3. Leadership, and the organization as a whole, MUST proactively protect against the misuse of telemetry.
  4. The gathering and use of telemetry MUST be excessively transparent1 to all human systems impacted by the telemetry.
  5. Human evaluations and compensation MUST NOT rely on the telemetry data listed within OpenO11y.
  6. When observing human systems, observers MUST continuously reevaluate the health and safety of the people and culture.
  7. Secondary contributors MUST NOT use telemetry to measure the teams they support in any way that violates these principles.
  8. Groups within a human system MUST NOT be compared to others in any way that violates these principles.
  9. You MUST NOT be an A$$hole2

Building A Culture of Being Safely Observed

Humans don't enjoy being watched, especially when observations are misused or directly used for measuring performance. Misusing telemetry can have a net negative effect, eroding trust in human systems and the individuals they are comprised of.

Tell me how you measure me and I will tell you how I will behave. If you measure me in an illogical way, do not complain about illogical behavior
- Eli Goldratt4

When observing human systems, it is critical to build a culture where individuals can both feel and be safe while being observed.

Employee Evaluations and Compensation

Being safely observed requires that employers MUST NOT use telemetry for reviews, compensation adjustments, employment decisions, etc. Because reviews and compensation are so impactful to employees, the use of telemetry data can dramatically shift an employee's focus to optimizing telemetry data. If employers misuse telemetry, then employees are negatively impacted and using telemetry to improve becomes unimportant. While it MAY seem like a positive change, optimizing telemetry often results in employees being less effective contributors. This is a problem that can be avoided.

Individual Telemetry

Telemetry data MUST only contain aggregated data for systems (groups) of people. Observability over human systems MUST NOT gather telemetry from specific individuals. This is a foundational requirement of being safely observed.

Comparing Disparate Groups

Being safely observed also extends to cross-group evaluations. Every team has distinct members, work responsibilities, personal responsibilities, opinions, etc. Every human system is unique. Thus, some of the telemetry produced by observing any two teams will differ naturally.

The Agile Framework provides a useful example with the nature of story points in software development. As a unit of measurement, the value of a story point is defined by each team. Thus, giving story points are relative, comparing story points across teams isn't valuable.

This is not inherently a problem. Yet, it does require us to be selective when comparing data across disparate groups. Considering the story points example, it is unreasonable to compare story points "completed" over a given time period between multiple teams. However, it MAY be reasonable to compare the story points planned / story points completed between multiple teams as an unusually low or high ratio could indicate a planning and/or capacity issue.

Don't ask people to collaborate if they know that, in the end, there will be a winner and a loser
- Bertrand Duperrin5


Using Telemetry In Different Roles

Teams and Individual Contributors

These are the ideal candidates for using telemetry data. Teams and individuals SHOULD observe themselves to identify trends and establish their working norms. Once a baseline is established, teams SHOULD then use telemetry to identify issues and strive to improve individual and team performance based on an established set of engineering principles.

Secondary Contributors

OpenO11y uses this term to describe roles that exist to organize, improve, or facilitate the work of others. In the world of software development, these roles commonly include, but are not limited to, SCRUM Masters, Project Managers, Product Owners, and Management. Secondary contributors SHOULD NOT use telemetry to observe the teams they support. Doing so offers limited benefits but can significantly damage the the organizations culture of being safely observed.

Senior Leadership MAY use this telemetry to solve a particular business case in the context of organizational transformation. Senior Leadership MUST ensure the use of this telemetry is done in a manner that promotes a safe working environment and culture. See Gathering and Using Telemetry for specifications on cultural use of telemetry.

Senior Leadership MAY involve lower levels of leadership and management in the above exception, particularly with respect to Direct Managers whose direct reports are individual contributors. These leaders MUST be especially careful with how they use telemetry data. These team members MUST constantly evaluate if they are observing their team or themselves as a part of that team.

All leaders MUST pay careful attention to reactions, team feedback, and changes in team behaviors.


Gathering And Using Telemetry

When gathering and using telemetry, organizations MUST must respect the Guiding Principles. The following list of actions can help ensure that leaders do not compromise the trust and safety of their organization while gathering and using telemetry data.

  1. Transparency in Data Collection: Leadership MUST clearly advertise what telemetry data is being collected and why. The purpose and benefits of the data collection MUST be communicated openly to all groups.
  2. Purposeful Use: Telemetry data MUST only be used for the openly communicated purposes. These purposes MAY be aligned with organizational goals such as improving productivity, identifying bottlenecks, enhancing efficiency, and fostering a culture of continuous improvement. However, the use of this telemetry data MUST NOT violate any Guiding Principles.
  3. Data Privacy and Security: The implementing organization MUST implement robust data security measures to protect telemetry data and respect privacy. This data MUST be anonymized and MUST NOT contain any PII3.
  4. No Misuse: Telemetry data MUST NOT be used for punitive measures. Instead, it SHOULD be used for constructive feedback, learning, and growth.
  5. Feedback Mechanism: Leadership MUST provide opportunities for team members to share their concerns or suggestions about how telemetry is collected and used. This helps in building trust and making improvements based on feedback.

  1. Transparent - Characterized by visibility or accessibility of information especially concerning business practices. Merriam Webster - Transparent 

  2. The No A$$hole Rule 

  3. Personal Identifiable Information 

  4. Eliyahu Goldratt 

  5. Bertrand Duperrin