Rose Jones · An engineering walkthrough

Agile Software
Development

Scrum for engineers who have mostly experienced it as meetings. Two decades of accumulated practice, reduced to the parts that make it actually work.

Scrum is a discipline, not a ritual. The ceremonies are the easy part; the contracts are the work.

01 · Why bother

The same ceremonies, two different teams

Ritual scrum

  • Standup is status theater for the manager
  • Estimates drift mid-sprint until they always "fit"
  • Stories carry over silently, sprint after sprint
  • Stakeholders ask "when will it be done?" daily, because the board answers nothing

Disciplined scrum

  • Standup surfaces blockers the team fixes that day
  • Estimates freeze at commitment; misses become lessons
  • Carryover triggers a retrospective question, not a shrug
  • Velocity gives a forecast; the board answers "when" so people stop asking

Predictability is the product of discipline. Everything in this deck exists to make one number trustworthy: how much this team ships per sprint.

02 · The shape of it

The sprint loop

1PlanningThe team signs up for a committed amount of work.
2Stand-upDaily. Surface blockers, not status.
3RefinementEstimate and clarify what enters the next sprint.
4DemoShow working software to stakeholders.
5RetrospectiveInspect the process; pick one thing to change.
1 2 3 4 5 Planning Stand-up, daily Refinement Demo Retrospective the sprint

Every ceremony feeds the next. Skip one and the loop leaks: no refinement, bad estimates; no retro, repeated mistakes.

03 · The contracts

The working agreement: five promises

1Speak upAnyone can amend the contract during refinement. After that, it holds.
2Action itemA story is immediately actionable: acceptance criteria written, dependencies clarified, requirements sufficient.
3Sprint commitmentWe, as a team, sign up for the committed amount of work. Not individuals: the team.
4CompletionA ticket is complete when it meets its acceptance criteria, the Definition of Done. Not before.
5DeliveryAt sprint end we answer honestly: how much of our commitment did we complete?

Write it down and make it visible. An unwritten agreement is just a collection of assumptions that disagree.

04 · Estimation

Two currencies: story points and earned value

Story points · engineers price the cost

Resource consumption: how much of the sprint must be reserved? Time plus uncertainty: risks, dependencies, unknowns, thin requirements.

CSV export: 5 points. The parsing is easy; the unknown load on the reporting database is not.

Earned value · stakeholders price the worth

Product and business answer one question: how much would I pay for this work? Value is their call, never the engineers' guess. Pricing worth also incentivizes finishing: value is earned at done, not at almost.

CSV export: EV 8. Three enterprise clients asked for it this quarter.

Easy win do these first Big bet schedule deliberately Backlog filler when convenient Poor use of resources question these CSV export · EV 8, 5 pts earned value story points

Priority is the ratio. High value at low cost goes first; high cost at low value gets questioned, however interesting it looks.

04b · Estimation

Can't estimate it? Spike it

A spike is a time-boxed investigation with its own acceptance criteria. It buys the information an estimate needs. The CSV ticket's unknown, spiked:

1QuestionCan the reporting database serve a 10k-row export live, or do we need a replica?
2Time boxTwo days, in hours or days, never "until we know". The clock is the scope.
3ArtifactNamed upfront in the acceptance criteria: a sample query plus a one-page write-up. A demo, video, or sample code also counts.
4Out of time?Report back anyway. If more investigation is needed, refine a follow-up spike; never let the clock quietly extend.

A spike that ends with "we need more time" and no artifact wasn't a spike; it was a delay with a nicer name.

05 · The gates

Definition of Ready, Definition of Done

Ready · may it enter the sprint?

Requirements
Export the current report view as CSV; columns match the on-screen table
Priority
EV 8 · 5 story points
Acceptance
10k rows download in under 5 s; commas and unicode escaped
Dependencies
Read replica confirmed by spike; no external teams
Scope
Excel formatting explicitly out

Done · may it be closed?

Code
Merged; review comments resolved; follow-up PRs merged too
Docs
Documentation updated; knowledge shared with the team
Production
Deployed and verified with a production run
Questions
Every question listed in the ticket answered
Follow-ups
New work becomes new stories, never silent scope creep

Ready gates entry; Done gates exit. Everything between the gates is the team's promise, and the estimate only means something if both gates hold.

06 · The payoff

Velocity: the number discipline buys you

  • Velocity is the historical average of points completed per sprint
  • Estimates freeze once a story enters the sprint; no re-pricing mid-flight
  • Stories close only through the Definition of Done
  • Garbage in, garbage out: undisciplined points make a meaningless average
  • Forecast: a 100-point epic at velocity 34 is three sprints, and you can say so out loud
283140 303639 s1s2s3 s4s5s6 average 34

A story carried over three sprints is not a failure to punish; it is a question for the retro: "what can we learn from this one?"

07 · Visibility

Automate the answers

One source of truth

The board and its timeline, generated from tickets the team already maintains. No parallel decks in Miro, Lucid, or spreadsheets; no duplicated content drifting out of date.

Public by default

Status is visible to anyone who wonders, including stakeholders. Timeline views roll sprints into a roadmap without a single status meeting.

Dev time back

Every automated answer is a status-check question nobody asks an engineer. The team's attention stays on the sprint, not on reporting about it.

If the board is the truth, no one has to ask. If people keep asking, the board isn't the truth yet.

08 · Takeaways

If you remember four things

1Contracts, writtenThe working agreement is five explicit promises, amendable only in refinement.
2Two currenciesEngineers price cost in points; stakeholders price worth in earned value. Priority is the ratio.
3Guard the gatesReady gates entry, Done gates exit. Spike the unknowns before they enter.
4Velocity is earnedFrozen estimates and honest closes make the average forecastable. Carryovers go to the retro.

The goal, in one line: a team whose word about "when" can be trusted.