Skip to content
Tutorials Center
My learning

Topics

Lesson 10 of 29 ยท Requirements and Estimation

Estimating with Story Points

Learn relative estimation with story points: the Fibonacci scale, planning poker, reference stories, capacity and how to forecast with velocity.

Sooner or later someone asks, "How long will this take?" People are surprisingly bad at predicting hours for work they have never done, but they are quite good at comparing: "this is about twice as hard as that". Story points use that strength. They are a unit of relative size that combines effort, complexity, uncertainty and risk into one number, so that the team can compare stories without pretending to know exact hours.

This lesson explains how relative sizing works, how the Fibonacci scale and planning poker keep estimates honest, how reference stories give the numbers meaning, and how a team turns points into a forecast. You will need user stories to estimate, so User Stories and Acceptance Criteria is a good prerequisite. The examples are JavaScript simulations: a stand-in for a planning session and a stand-in for forecasting. Real estimation is a conversation among people, and no script replaces it.

Why points, not hours?

Relative Compare stories, don't guess time.
Uncertainty Bigger numbers admit bigger doubt.
Team-owned Your points are not my points.

Points are a tool for planning, not a promise.

1 / 2

What a story point measures

A story point is not a unit of time. A "5" does not mean five hours or five days. It means "bigger than a 3, smaller than an 8", and the team decides what that feels like. Three things feed the number:

  • Effort: how much work there is to do.
  • Complexity: how hard or tricky the work is.
  • Uncertainty and risk: how much is unknown, such as a new library or unclear rules.

A small change in an unknown part of the system can be larger than a big change in code you know well. Points capture that, which hours rarely do.

Points are also team-specific. A team of experts and a team of newcomers will give the same story different numbers, and both can be right, because each team's points are only meaningful against its own other stories. Never compare story points across teams, and never use them to compare people. Both habits turn a planning aid into a scoreboard (see Velocity, Burndown and Cumulative Flow for what goes wrong).

Relative sizing works for a reason: humans estimate ratios far better than absolute values. Most of us cannot say how long a task will take, but can say that this task is clearly bigger than that one. The team keeps estimating until there is enough for a plan, no more.

The Fibonacci scale

Most teams use a scale that looks like the Fibonacci sequence: 1, 2, 3, 5, 8, 13, 21, sometimes with 0, a half point, 40, 100 and a "?" or a coffee cup card for "I need a break" or "I can't estimate this".

Why not 1, 2, 3, 4, 5, 6...? Because the gaps grow as the numbers do. The difference between 1 and 2 is easy to judge, while the difference between a 14 and a 15 is a false precision no one can defend. The widening gaps force you to choose a bucket: if you are torn between 5 and 8, you must pick one rather than argue over a 6 or 7. They also reflect that bigger work carries more uncertainty.

A few widely used conventions:

  • Anything bigger than the team's largest comfortable number (often 13 or 21) is too big and should be split before it enters a Sprint.
  • A story of 1 or 2 points is small enough that sub-dividing it further is not worth the effort.
  • T-shirt sizes (S, M, L, XL) work in the same way. Some teams use them for early, rough sizing and switch to points once stories are close to being built.

Reference stories

Numbers mean nothing without an anchor. A reference story (or baseline) is a past story that the team remembers well and agrees on a point value for. When a new story comes up, the question is not "how many points is this?" but "is this bigger or smaller than our reference 3, and by how much?".

A new team can start by choosing a small, well-understood story and calling it a 2 or a 3, and then sizing the others around it. Keep two or three references of different sizes on the wall or in the backlog tool, for instance:

Points Reference story
2 Add a field to the booking form.
5 Cancel a booking from the email link.
13 Take card payments for a deposit.

Revisit the references once in a while. If the team's understanding drifts, the numbers drift with it, and old data stops being useful for forecasting.

Planning poker

Planning poker is a popular technique for estimating together, and it is designed to defeat anchoring: the tendency to adjust toward the first number you hear. The steps are simple.

  1. The Product Owner reads the story and answers questions.
  2. Each Developer privately chooses a card from the scale.
  3. Everyone reveals at the same time.
  4. If the votes agree, that is the estimate. If they differ widely, the highest and lowest voters explain their thinking, and the team votes again.

The most useful moment is the disagreement. A 2 next to a 13 usually means the two people imagine different scopes, or one knows a hidden problem the other does not. The conversation uncovers assumptions, and the estimate often settles after a round or two. If it does not, the story probably needs splitting or a short investigation first, called a spike.

The Product Owner takes part in the discussion but does not vote. The Developers, who do the work, own the estimate.

A planning poker simulation

The stand-in below simulates one session. Four invented developers have slightly different tendencies (Ana is optimistic, Chen cautious), and their first votes have a lot of noise, as real first votes do. A seeded random number generator makes the run repeatable. After each reveal, the code checks whether the lowest and highest cards are more than one step apart on the scale; if so, a discussion happens and the next round is less noisy.

JS Simulate planning poker

Press Run (or Ctrl + Enter) to see the console output here.

No output.

In the first round Ana votes 2 while the others vote 8, a wide gap, so the team discusses before voting again. In the second round the votes land on 3 and 5 and the team settles on 5. Try changing the seed or the story's trueSize and watch how the number of rounds changes. The model is only a toy: it can show the shape of the process, not the real value of the conversation.

Capacity and velocity

Once stories have points, the team can ask how many fit in a Sprint. There are two common answers.

Velocity is the number of points the team completed in recent Sprints. Only items that met the Definition of Done count, never partial credit. After three or four Sprints, the average velocity is a reasonable guess for the next one.

Capacity is a bottom-up view: how many people are available, for how many days, after holidays, leave and meetings. Teams sometimes plan in hours or days for the task breakdown and use points only for the backlog. Both views are useful: velocity says what the team usually does, capacity says what is different this time. If two developers are on holiday, expect less than the average.

Velocity is a forecast, not a target. It describes the past, it is noisy, and it changes when the team, the product or the way of working changes. Teams that are pushed to raise it tend to inflate their estimates, which makes the number useless.

Forecasting with velocity

With a stable velocity history, forecasting is simple division. If the remaining backlog is 130 points and the team averages about 23 points per Sprint, expect around six Sprints. Because velocity varies, always give a range, using the best and worst recent Sprints, not one date.

JS Forecast sprints from velocity

Press Run (or Ctrl + Enter) to see the console output here.

No output.

The output gives a range of 12 to 14 weeks, not a single date, which is the honest way to share a forecast. The second half plans one Sprint by taking backlog items in priority order until the likely velocity is used up, skipping the 13-point story because it does not fit. A real team would also check the Sprint Goal and might split the 13-point story instead of skipping it. Compare this with the fuller version in Project: Sprint Planner.

Do not turn points into hours (for example, "1 point = 4 hours") or use them to judge individuals. Once points become a conversion rate or a performance target, the team stops estimating honestly.

Common mistakes

  • Estimating alone. The learning happens when different people compare their views.
  • Debating for ages over a 5 versus an 8. If the votes are adjacent, take the higher one and move on.
  • Re-estimating finished stories to make velocity look better.
  • Estimating everything in the backlog. Size only what is likely to be built in the next few Sprints.
  • Treating the estimate as a commitment. It is a forecast that gets updated when you learn more.
  • Ignoring unfinished work. Partially done stories do not add to velocity.

Some teams skip estimation altogether and just count stories (a "no estimates" or throughput approach, covered in Kanban and WIP Limits), which works well when stories are small and similar in size. Story points are a tool, not a rule of Scrum; the Scrum Guide mentions sizing but does not require points.

Check your understanding

Quick check ยท 5 questions

of 5 answered You scored / 5 ยท perfect!

  1. 1. What does a story point primarily measure?

    Story points express relative size. They are not hours, and they do not measure value. Effort, complexity and uncertainty all feed the number.
  2. 2. Why do planning poker players reveal their cards at the same time?

    Simultaneous reveal stops people from adjusting their vote to match the first (or most senior) voice, so the discussion starts from independent opinions.
  3. 3. Why does the usual scale use 1, 2, 3, 5, 8, 13 instead of every whole number?

    As work gets bigger, the estimate gets less certain, so the scale spaces numbers further apart and forces you to pick a bucket.
  4. 4. A team's last three Sprints finished 20, 24 and 22 points. About how many points should it plan for the next Sprint, assuming nothing changes?

    The average of recent Sprints, here 22, is a reasonable forecast. It is not a target to raise, and it should be adjusted for known changes in capacity.
  5. 5. Which use of story points is a bad practice?

    Points are team-specific, so comparing velocity across teams has no meaning and pushes teams to inflate estimates.

Summary

  • Story points measure relative size, combining effort, complexity and uncertainty. They are not hours.
  • The Fibonacci-like scale (1, 2, 3, 5, 8, 13, 21) widens the gaps as stories get bigger, which avoids false precision.
  • Reference stories give the numbers meaning for your team. Points are not comparable between teams.
  • Planning poker reveals independent opinions and uses disagreement to surface hidden assumptions.
  • Velocity is the points completed per Sprint. Use recent history and capacity to forecast a range, not a date.
  • Never use points as a target, a conversion to hours or a way to rank people.

Next, look at work that flows continuously instead of in Sprints in Kanban and WIP Limits.