Guide

Sprint Estimation for Remote Teams

How remote scrum teams run sprint estimation that works: deck choice, facilitation scripts, timeboxing, and no-signup voting for guests.

What sprint estimation actually is

Sprint estimation is blind, simultaneous, relative estimation. Every voter picks a card for a story at the same time, all cards flip together, and the spread — not the average — is the useful signal. A room that votes 3, 3, 3, 13 has not found a number; it has found a misunderstanding, and that is exactly what you want to surface before the sprint starts.

You will see the same practice called story point estimation, sprint estimation, scrum estimation, or story point sizing. The names are interchangeable. What differs between teams is discipline: deck, timebox, and who is allowed to speak first.

Why remote estimation breaks

• Anchoring. On a video call the loudest or most senior voice says "that's a five" and the room converges on it. Blind voting is the only reliable fix.
• Silent stalling. Without a visible timer, a single distracted participant holds the whole round hostage.
• Tool friction. If your estimation tool asks eight people to create accounts, you lose the first five minutes of every session and at least one contractor entirely.
• Invisible context. Remote teams cannot point at a whiteboard, so the story under discussion must stay on screen at all times.
• Observers voting. Product managers and stakeholders skew the numbers. They should watch, not vote.

Fibonacci vs. modified Fibonacci

Standard Fibonacci (1, 2, 3, 5, 8, 13, 21) grows fast enough that the gaps themselves communicate uncertainty — you cannot pretend to know the difference between a 13 and a 14. Modified Fibonacci (0.5, 1, 2, 3, 5, 8, 13, 20, 40, 100) adds a half point for trivial work and coarse buckets for epics that should be split rather than estimated.

Pick one deck and keep it for at least three sprints. Velocity is only meaningful if the scale stops moving. Add a ? card so people can say "I don't understand this yet" instead of guessing, and treat any ? as a blocker for that story.

A 45-minute remote session that works

1. 0–5 min — open the room. Host creates the session and drops one link in the team channel. Everyone joins with a display name.
2. 5–8 min — calibrate. Agree on one reference story the whole team knows and call it a 3. Every later estimate is relative to it.
3. Per story — 60 seconds of context. The product owner reads the story and acceptance criteria. No solutioning yet.
4. 30 seconds of silent voting. Timer visible to everyone. Cards stay hidden.
5. Reveal and read the spread. If the votes are within one card, take the higher number and move on — do not discuss.
6. Two minutes for outliers only. Ask the lowest and highest voters to explain, in that order. Then re-vote once.
7. Split, don't debate. A story that fails two rounds is too big or too vague. Split it and estimate the pieces next time.

Facilitation rules worth enforcing

• Lowest voter speaks before the highest voter, and both before anyone else.
• Estimate effort and risk, never hours. Points are not a timesheet.
• Only the people who will do the work vote; everyone else is an observer.
• Re-vote at most twice per story, then split it.
• Reset the round automatically whenever the story changes, so stale votes never leak.

Removing the signup wall

The highest-leverage change most distributed teams can make is not a better deck — it is removing the account requirement. Stacked is built around that: the host signs in with Google or email, creates a session, and shares a single URL. Everyone else — teammates, contractors, a stakeholder who wants to watch — types a name and is voting within seconds. No invites, no seats, no waiting.

The rest follows the rules above: Fibonacci and modified Fibonacci decks, a host-set round timer defaulting to 30 seconds, votes held server-side until the host reveals, observer mode for non-voters, and automatic round resets when the story under estimation changes.

Frequently asked questions