📋 Backlog Grooming

Estimate Story Points via Planning Poker

Facilitates a structured planning poker session by presenting each story with discussion questions, estimation prompts, and a variance analysis. Useful for distributed Scrum teams estimating asynchronously.

This prompt runs a planning poker session using the modified Fibonacci scale (1, 2, 3, 5, 8, 13, 21) with explicit variance analysis when estimates span more than two adjacent values. It generates discussion questions for each outlier to resolve before a re-vote.

When to use this prompt

Reach for this when your team does async estimation and you need a facilitator-style script to walk through 8-15 stories without a live meeting. You will need the story descriptions and any past-velocity data you want the model to use for calibration. It assumes your team already uses the modified Fibonacci scale; if you use t-shirt sizes, adapt the values. The prompt shines when there is real disagreement on estimates because it generates the discussion prompts needed to converge, rather than just averaging votes.

The Prompt

Role: Scrum Master Variables: {{team_size}}, {{team_velocity}}, {{reference_story}}, {{reference_points}}, {{stories}}

You are an experienced agile coach facilitating a planning poker session for {{team_size}} engineers. I will give you stories to estimate. Your job is to walk the team through each one with a consistent structure.

Estimation scale: modified Fibonacci (1, 2, 3, 5, 8, 13, 21)
Team velocity: {{team_velocity}} points per sprint
Reference story: {{reference_story}} = {{reference_points}} points

Stories to estimate:
{{stories}}

For each story, produce:

1. STORY PRESENTATION — 2-3 sentence restatement, plus 1 key unknown to discuss.

2. DISCUSSION PROMPTS — 3 questions the team should answer before voting (e.g., "Is the API contract stable?" "Have we estimated something like this before?").

3. ESTIMATE ANCHOR — Suggest a likely range based on the reference story (e.g., "Probably 3-5 based on similarity to reference").

4. VARIANCE PROTOCOL — If estimates span more than two adjacent Fibonacci values (e.g., someone says 3 and someone says 13), the outliers explain their reasoning and the team re-votes. Write the specific question for the outliers in each case.

At the end, produce a "Convergence summary" that tells the Scrum Master which stories likely need the most discussion time. Never average votes; always re-vote after discussion.

Example Output

STORY 1: Add OAuth2 login for Google
PRESENTATION: Users authenticate via Google instead of email+password. Key unknown: do we store a refresh token or re-auth on each session?
DISCUSSION PROMPTS:
- Is the OAuth library already in the codebase?
- Do we need to handle existing email users migrating?
- Who owns the Google Cloud project?
ESTIMATE ANCHOR: Likely 5-8 if the library exists; 13 if greenfield.
VARIANCE PROTOCOL: If range spans 3 to 13, ask low voters if they assumed the library exists, and high voters if they assumed greenfield infra work. Re-vote after alignment.

Convergence summary: Stories 1 and 4 have the highest expected variance (both touch auth). Allocate 6-8 minutes each. Stories 2 and 3 should converge quickly; budget 3 minutes.

Recommended Tools

Linear and Jira have built-in story point fields that capture estimates directly on the issue, so the output of each round can be logged without switching tools. Taskade supports collaborative async voting with a live view, which helps distributed teams run planning poker without a video call. All three support the Fibonacci scale natively or via plugins.

Linear74/100 Jira Software75/100 Taskade65/100

Frequently Asked Questions

When should I use this prompt?

Use it for async estimation with distributed teams, or as a prep document before a live planning poker meeting so people show up already primed. It is most valuable on teams where estimation debates regularly eat the entire session. If your team already converges within 5 minutes per story and has stable velocity, the structure is unnecessary overhead. Skip it for tiny tickets (under 2 points by eye) and reserve for the medium-to-large stories where disagreement costs the most.

What if my team uses t-shirt sizes instead of Fibonacci?

Replace the modified Fibonacci line in the prompt with your scale (XS, S, M, L, XL, XXL) and adjust the variance protocol to trigger when estimates differ by more than one size. T-shirt sizing is less precise but lower ceremony, so you may also want to drop the "estimate anchor" step since reference stories matter less. The discussion prompts and convergence summary work unchanged.