Skip to content

01 · Agile Project Management

Level 1 described the predictive (waterfall) lifecycle: define scope up front, plan the whole schedule, then execute against it. Agile inverts the assumption. Instead of treating scope as fixed and letting cost and schedule absorb uncertainty, agile fixes cost and schedule (a team of a known size, working in fixed-length iterations) and lets scope flex — you deliver the highest-value slice of the backlog that fits, and re-decide what's next every iteration. This module covers the two dominant agile frameworks a PM actually encounters — Scrum and Kanban — plus the estimation and forecasting math that makes an agile plan defensible to a sponsor who wants a date.

When agile fits, and when it doesn't

Agile is not universally superior; it's a fit for a particular kind of uncertainty. The practical test is: is the requirement stable enough to plan in detail six months out?

Condition Predictive fits Agile fits
Requirements clarity Well understood, stable Emerging, expected to change
Feedback loop Value only visible at the end Partial product is usable and testable early
Regulatory/contract constraint Fixed scope contractually mandated Scope negotiable within a fixed budget
Cost of change late Very high (physical build, hardware) Low (software, content, service design)
Stakeholder availability Sign-off at gates Continuous, weekly involvement

Most real portfolios are hybrid: an agile delivery team inside a predictive governance wrapper (fixed annual budget, stage-gate funding, milestone reporting to a steering committee). Knowing how to translate between the two — velocity into a forecast date, story points into a spend rate — is the core PM skill in this module.

Scrum: roles, events, artifacts

Scrum is a lightweight framework built around fixed-length iterations called sprints (typically 2 weeks).

Element Type Purpose
Product Owner Role Owns and orders the product backlog; single voice on priority and acceptance
Scrum Master Role Facilitates the process, removes impediments, protects the team from mid-sprint churn
Developers Role The people who build the increment; self-managing on how the work gets done
Sprint Planning Event Team pulls the top of the backlog into a sprint goal and sprint backlog
Daily Scrum Event 15-minute team sync on progress toward the sprint goal and blockers
Sprint Review Event Demo the increment to stakeholders; gather feedback that reshapes the backlog
Sprint Retrospective Event Team inspects its own process and commits to one or two concrete improvements
Product Backlog Artifact Ordered list of everything wanted in the product
Sprint Backlog Artifact The subset pulled into this sprint, plus the plan to deliver it
Increment Artifact The working, potentially releasable output of the sprint

Two rules do most of the work: the sprint goal doesn't change mid-sprint (new requests go to the backlog for the next planning session), and the increment must meet the team's Definition of Done — a standing checklist of what "done" means (code reviewed, tests passing, documentation updated, deployed to staging). Without a Definition of Done, velocity is measuring partially-finished work, and every forecast built on it is fiction.

Kanban: flow instead of iterations

Kanban has no sprints. Work moves continuously across a board, and the control mechanism is a work-in-progress (WIP) limit on each column.

Backlog Ready (WIP 5) In Progress (WIP 3) Review (WIP 2) Done
40+ items 4 items 3 items 2 items

When "In Progress" is at its limit of 3, nobody may start a fourth item — the team must first help finish something already in flight. That is the entire point: WIP limits convert a team's instinct to start work into an instinct to finish work, which is what actually reduces cycle time.

Metric Definition What it tells you
Lead time Request accepted → delivered What a stakeholder experiences
Cycle time Work started → delivered The team's actual throughput speed
Throughput Items completed per week Capacity, for forecasting
WIP Items in flight right now Early warning — rising WIP predicts rising cycle time
Scrum Kanban
Cadence Fixed sprints Continuous flow
Commitment Sprint goal per iteration Per-item, pull-based
Change mid-cycle Discouraged Allowed any time, respecting WIP
Core metric Velocity Cycle time / throughput
Best for Feature development against a roadmap Support, ops, maintenance, unpredictable arrival of work

Estimation: story points and capacity

Story points are a relative measure of size (complexity + effort + uncertainty), not hours. They work because humans compare well ("this is about twice that") and forecast absolute durations badly. Teams typically estimate on a modified Fibonacci scale — 1, 2, 3, 5, 8, 13 — where the widening gaps encode the fact that big things are estimated less precisely.

Capacity is the separate, concrete question of how many person-hours the sprint actually contains. For a 2-week (10 working day) sprint:

Person Days available Productive hours (6/day)
Dev 1 10 60
Dev 2 8 (2 days PTO) 48
Dev 3 10 60
QA 9 (1 day training) 54
Designer 7 (3 days on another project) 42
Product Owner (part-time) 5 30
Total 294 hours

Note the six-productive-hours-per-day assumption: an 8-hour day contains meetings, email, and interruptions. Applying a focus factor of 0.8 to absorb unplanned support work gives 294 × 0.8 ≈ 235 hours of real sprint capacity. A team that plans against 6 people × 10 days × 8 hours = 480 hours will over-commit by roughly a factor of two and miss every sprint.

A worked example: forecasting a release date from velocity

A team has completed five sprints. Velocity (story points accepted, meeting the Definition of Done) per sprint:

Sprint 1 2 3 4 5
Velocity 23 31 28 34 29

Average velocity = (23 + 31 + 28 + 34 + 29) ÷ 5 = 145 ÷ 5 = 29 points per sprint. The remaining backlog for the release is estimated at 260 points.

  • Expected case (average velocity 29): 260 ÷ 29 = 8.97 → 9 sprints
  • Optimistic case (best observed, 34): 260 ÷ 34 = 7.65 → 8 sprints
  • Pessimistic case (worst observed, 23): 260 ÷ 23 = 11.30 → 12 sprints

With 2-week sprints, that is a forecast range of 16 to 24 weeks, most likely 18 weeks. This is the correct way to answer "when will it be done?" in an agile context: a range grounded in the team's own measured throughput, not a single date grounded in optimism. Reporting only the 18-week figure hides the fact that a run of sprints at the low end pushes the release out by a further six weeks.

Two things invalidate this forecast, and watching for both is the PM's job. First, scope growth: if the backlog grows by 40 points during those sprints, the expected case becomes (260 + 40) ÷ 29 = 10.34 → 11 sprints, so the forecast must be recalculated every sprint against the current backlog, not the original one. Second, an inconsistent Definition of Done: if sprint 4's 34 points included two stories that were demoed but never actually tested, the real velocity was closer to 26, and the whole forecast rests on an inflated number. Velocity is a forecasting instrument only when "done" means the same thing every sprint.

Reporting agile work to a predictive governance layer

A steering committee that funds by quarter does not want story points. The translation table below is what a PM prepares before that meeting:

Sponsor's question Agile evidence Translation
When will it be done? Velocity 29, range 23–34; backlog 260 points "18 weeks expected, 16–24 week range"
What will it cost? Budget ÷ total points = cost per point Remaining points × cost per point
What are we getting? Sprint review demo A working increment, not a status slide
Is scope under control? Backlog size trend per sprint A rising backlog means scope growth — flag it
Are we on track? Velocity trend plus burndown slope Declining velocity is the leading indicator

Velocity is a planning tool, not a performance target

The moment velocity becomes a number the team is judged on, it inflates: estimates drift upward, the same work gets scored 8 instead of 5, and the metric stops predicting anything. Compare a team only against its own history, never against another team's points, and never set a velocity target in a performance review.

How It Actually Works

Velocity-based forecasting works because story points, properly estimated, converge to a roughly normal distribution around the team's true throughput once enough sprints are sampled — which is exactly why forecasting off one sprint's velocity is unreliable (n=1, no variance estimate) while forecasting off a trailing average of the last 6 sprints is far more stable: Sprints remaining = Remaining points / mean(last 6 velocities), and the standard deviation of those 6 velocities gives you a confidence range the same way PERT's σ does in Level 2 Module 2 — a mean velocity of 32 with σ = 4 means an 84%-confidence forecast should use 28 points/sprint, not 32. Kanban's flow metrics rest on the same Little's Law identity introduced in Level 1 Module 9: Cycle time = WIP / Throughput, which is why a Cumulative Flow Diagram's band widths (not heights) are the real signal — a widening "In Progress" band means WIP is growing faster than throughput, and the cycle-time increase it predicts will show up 1–2 weeks later even though today's board still looks fine. Story-point estimation itself works statistically only because relative sizing (Fibonacci-like scales: 1, 2, 3, 5, 8, 13) forces estimators away from false precision — the gaps between values widen with size specifically because uncertainty about a 13-point item is genuinely larger in absolute terms than uncertainty about a 2-point item, even if both feel equally "estimated."

Exercise

Take a project you know and set it up as an agile delivery. (1) Write a product backlog of at least 12 items, estimated in story points on the 1/2/3/5/8/13 scale, and state the relative reasoning for two of them ("this is a 5 because it's about twice the 3 we did last sprint"). (2) Build a capacity table for a 2-week sprint with a realistic team, applying a productive-hours assumption and a focus factor — show your arithmetic. (3) Write a Definition of Done with at least five checkable criteria. (4) Invent five sprints of plausible velocity, compute the average, and produce a three-point release forecast (optimistic / expected / pessimistic) in sprints and in weeks for your remaining backlog. (5) Finally, state whether Scrum or Kanban is the better fit for your project, and defend the choice against the "when agile fits" table above.