Skip to content
Tutorials Center
My learning

Topics

Lesson 28 of 50 ยท Agile and Teamwork

Estimation and Planning

Learn how teams estimate software: story points, planning poker, velocity forecasting, the cone of uncertainty and roadmaps, with runnable calculators.

"How long will it take?" is the question every software team dreads. Estimating is hard because software work is new each time: you are rarely building the same thing twice, and you only discover many requirements while building. Yet teams need some answer to decide what to start, what to promise and when to talk to customers.

The honest view is that an estimate is a forecast with a range, not a promise. Good estimation practice is mostly about producing useful ranges cheaply, being honest about uncertainty, and improving the forecast as real data arrives. Planning and estimating are people activities, so the code below is plain JavaScript stand-ins with invented numbers; there is no real tool behind them.

Estimating size, not time

People are poor at predicting absolute durations ("this will take 3 days") but reasonably good at comparing things ("this is about twice as hard as that"). Agile teams exploit this by estimating relative size in story points.

A story point is an arbitrary unit that expresses the combined effort, complexity and uncertainty of a backlog item compared with other items. A 2-point story is roughly twice the work of a 1-point story. One team's point is not the same as another's, and a point is not an hour. This is intentional: it stops managers from converting points into hours and judging people.

The usual scale is a trimmed Fibonacci-like sequence: 1, 2, 3, 5, 8, 13, 21. The gaps grow because larger estimates are less precise. Arguing whether something is 12 or 13 is pointless; the scale forces you to pick 13. Teams first agree on a reference story ("filtering a list is a 3") and size everything else against it. An item that is too big for the top card (often 13 or 21) should be split into smaller stories.

Why relative sizing works

Compare People judge ratios better than hours.
Coarse scale 1, 2, 3, 5, 8, 13 avoids fake precision.
Split big ones A 21 is a story waiting to be cut.

Points measure size and uncertainty, not time.

1 / 2

Planning poker

Planning poker is a technique for estimating as a group without anchoring. The product owner explains a story and answers questions. Each developer privately picks a card from the deck. Everybody reveals at once. If the cards agree, you have an estimate. If they differ widely, the people with the highest and the lowest card explain their reasoning (the low voter may know a shortcut; the high voter may have spotted a hidden risk), and the team votes again. Two or three rounds usually converge.

Revealing at the same time matters. If the most senior engineer says "that's a 3" first, everyone else quietly adjusts. The value of the exercise is the conversation, which surfaces assumptions, not the number.

This stand-in applies a simple rule to three rounds: identical cards are consensus; neighbors on the deck are close enough (take the higher); anything wider triggers a conversation.

JS Planning poker rounds

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

No output.

From points to a forecast: velocity

Velocity is the number of story points a team completes per sprint, counting only items that meet the definition of done (see Scrum in practice). After a few sprints you have a history, and you can forecast: remaining points / velocity = sprints left.

Never use a single number as the forecast. Velocity varies from sprint to sprint, so use a range, for example from your worst three sprints to your best three, or an average of the last three to six. Also remember what velocity is not: it is not a productivity score, and comparing it across teams is meaningless, since each team's points are its own. Pushing for higher velocity just makes teams inflate their estimates.

JS Velocity forecaster

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

No output.

The honest message is "between 6 and 7 sprints, most likely 7". Rounding up is deliberate: you cannot finish 6.9 sprints and have a finished increment. When the scope also changes (new items arrive), forecast with the current backlog and update the forecast each sprint.

The cone of uncertainty

Early in a project, estimates are very uncertain, and they become more accurate as decisions are made and work is done. This idea is usually drawn as a cone of uncertainty: wide on the left, narrowing to the right. A commonly quoted version, often credited to Barry Boehm and Steve McConnell, gives these approximate multipliers around the most likely value:

Stage Low High
Initial concept 0.25x 4x
Requirements agreed 0.67x 1.5x
Design complete 0.8x 1.25x
Detailed design done 0.9x 1.1x

Treat these numbers as an illustration of the shape, not as measured law. Two lessons follow. First, an estimate given at the idea stage is only a rough bracket, so be suspicious of anyone who gives an exact date then. Second, the cone does not narrow by itself: it narrows only because uncertainty is removed by making decisions and doing the risky work early (see spiral and risk-driven development).

JS Cone of uncertainty

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

No output.

For clarity, the code applies the same 20-week figure at every stage. In real life the point estimate itself changes as you learn, and the multipliers are applied to the estimate made at that stage.

Roadmaps: planning at different horizons

Planning happens at several levels, and detail should shrink as the horizon grows:

  • Product roadmap: themes and goals for the next quarters or year. It is a direction, not a list of dates. Items are outcomes ("customers can pay by invoice"), and the farther out, the vaguer.
  • Release plan: which stories are likely to ship in the next few sprints, based on velocity.
  • Sprint plan: the specific work for the next sprint, with tasks.
  • Daily plan: the team's own coordination.

Planning in detail far into the future wastes effort, because things will change. A popular rule is now / next / later: well-specified items for now, rough ideas for next, themes for later. When a stakeholder needs a date, give a range with its assumptions ("probably March to April, if the payment provider's API is ready by February") and revisit it as risks disappear.

Common estimation traps

  • Anchoring: the first number spoken biases everyone. Reveal estimates together.
  • Planning fallacy: people underestimate even when they know similar tasks ran long before. Use history (velocity) over gut feeling.
  • Padding: each layer of management adds a safety margin, and the inflated total hides real risk. Keep a visible, shared buffer instead.
  • Estimates become commitments: once a number is written down it is treated as a deadline. Say "forecast" and show the range.
  • Estimating work you do not understand: if nobody can size a story, schedule a short, time-boxed spike to learn first.

Check your understanding

Quick check ยท 5 questions

of 5 answered You scored / 5 ยท perfect!

  1. 1. Why do many agile teams estimate in story points instead of hours?

    Relative sizing is quicker and often more reliable. Points are deliberately team-specific units, which also stops them being used as a performance metric.
  2. 2. In planning poker, the cards are Ana=3, Ben=3, Chen=13, Dara=3. What should happen next?

    A big outlier means someone knows or assumes something the others do not. The discussion is the main benefit of planning poker.
  3. 3. A team's recent velocities are 20, 24 and 22. 110 points remain. How many sprints does the typical velocity forecast?

    Average velocity is 22, and 110 / 22 = 5 sprints. Remember to report it as a range, not a single promise.
  4. 4. What does the cone of uncertainty say?

    At the idea stage estimates can be off by a large factor in either direction. The range narrows only if uncertainty is actively removed.
  5. 5. Which is the best answer when a stakeholder asks for a date for a project in an early stage?

    A forecast with a range and stated assumptions is honest and useful, and it can be refined as the team learns.

Summary

  • Estimates are forecasts with ranges, not promises.
  • Story points express relative size and uncertainty on a coarse scale (1, 2, 3, 5, 8, 13, 21); split anything too big.
  • Planning poker gets independent estimates and uses disagreement to surface hidden assumptions.
  • Velocity (points per sprint) turns remaining points into a forecast of sprints; use a range from history, not one number.
  • The cone of uncertainty shows estimates are widest early; they narrow as decisions are made and risk is retired.
  • Plan in detail only for the near future; keep roadmaps about outcomes and use now, next and later.

Next: Working in a Team: Reviews, Communication and Ownership.