AI summary by TwinMind

A better way to estimate

Theo - t3․gg

12m 12s

Problems with Traditional Estimation

  • Estimations extremely difficult - never worked on team good at them
  • Most teams terrible/okay at estimations at best
  • Traditional point systems often frustrate when tasks don't complete on time
  • Ping doesn't currently use points due to difficulty making valuable system

Goals of Estimation Systems

  • Identify who should do task
  • Identify when task should be done
  • Identify how long task will take
  • Prioritize based on above factors
  • Identify low performers/underutilized team members

Key Insight: Time Estimation Failure

  • "How long it will take" fundamentally unreliable
  • Could be accurate sometimes but not consistently
  • Common trap: assuming point value = guaranteed completion time
  • This assumption is fundamentally wrong

Proposed New Estimation Framework

Three-Dimensional Scoring (1-10 scale each):

Complexity (or Specialization)

  • How hard/specialized is the task
  • How skilled should person doing it be
  • Example: WebRTC task at Ping = 10/10 complexity (specialized knowledge required)

Effort (formerly Time)

  • How long should task be worked on
  • Not mapped to specific time units
  • 1/10 = quick one-line fix, 10/10 = extended work period
  • Will almost always be wrong but provides relative indication

Risk (Confidence)

  • Chances that complexity/effort estimates are wrong
  • Missing component in most estimation systems
  • Critical for indicating estimate reliability

Real Example: Voice Levels in Pre-call View

Initial Estimates:

  • Complexity: 3-4/10 (seemed simple, reuse existing code)
  • Effort: 2/10 (quick implementation)
  • Risk: 6/10 (third-party dependencies, device interaction)

Actual Results:

  • Agora SDK doesn't provide reliable voice indication outside calls
  • Risk estimate proved accurate
  • Complexity became 6-7/10, Effort became 4-5/10
  • High risk score correctly predicted estimate inaccuracy

Value of Multi-Dimensional Approach

Task Assignment Benefits:

  • Easier identification of right person for task
  • Clear reasoning for assignment decisions
  • Better employee development planning

Junior Developer Growth Strategy:

  • Start with: low complexity, high effort tasks
  • Progress to: medium effort, high complexity tasks
  • Consistently impressive results when shifting focus from effort to complexity
  • Reveals unexpected developer capabilities

Key Problem with Traditional Systems

  • Conflate complexity and effort into single metric
  • Ignore risk entirely
  • Pretend complexity = effort (fundamentally different concepts)

Handling Unknowns and Research

Separate Research Tasks:

  • When complexity/risk unclear, create research task first
  • Example: "Identify how to get voice levels from third-party SDK"
    • Complexity: 2/10 (read docs)
    • Effort: 2/10 (research task)
    • Risk: 4/10 (moderate uncertainty)
  • Research tasks mitigate risk for subsequent implementation
  • If struggling to categorize complexity vs risk = need research task

Current Implementation Status

Ping's Current Approach:

  • Team of 5 developers + 1 non-developer
  • Using gut feel due to small team size
  • All tracking in Linear
  • Good intuitive sense of task assignment and timing
  • Thinks in proposed framework but doesn't formally measure yet

Future Plans:

  • Will implement formal system when team scales
  • Proposal for future estimation methodology
  • Moving away from T-shirt sizes and Fibonacci sequences
  • Focus on: what's hard + who has skills, what's long + who has time, what's risky + how wrong are estimates

Implementation Philosophy

  • Estimation system should expect to be wrong
  • Less guarantee, more indicator for task assignment
  • Separate unknowns identification as distinct task type
  • Build system that adds value rather than frustration

TwinMind

Get summaries like this for your own meetings

TwinMind takes notes in the background during meetings, lectures, and calls, then writes the summary for you.

Free plan to get started

No bot joins your calls

Phone, desktop & Chrome apps

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome