Chapter 2 · Requirements
User Stories, Use Cases and Prioritising
- Page 7 of 18
- 17 min read
ClinicQueue now has written requirements and acceptance criteria. But a list of twenty "shall" statements is hard to plan with: which ones belong together, which can be built first, and which can wait? Agile teams usually break the work into user stories, small pieces of value told from a user's point of view, and keep them in one ordered list. Some situations need more detail than a story can hold, and there a use case helps.
On the previous page, Functional and Non-functional Requirements, and Acceptance Criteria, you learned to make requirements testable. This page turns them into stories and use cases, and ends with something the team will use for the rest of the project: a prioritised product backlog for ClinicQueue.
What you will learn
- The user story format and the three Cs: card, conversation, confirmation.
- How to check a story with the INVEST criteria, with weak and improved examples.
- How to split a large story (an epic) into small ones that each deliver value.
- How to write a use case with an actor, a main flow and alternative flows, and when a use case is better than a story.
- How to prioritise with MoSCoW and with value versus effort, and build an ordered product backlog.
User stories
A user story is a short description of something a user wants to do, and why. A common template, which Dan North shows as the story form in his article "Introducing BDD", is "As a [X] I want [Y] so that [Z]": X is the role, Y is the action, Z is the benefit.
Example:
US-01 As a patient, I want to book a free slot online,
so that I do not have to phone the clinic during its busiest hours.
US-04 As the receptionist, I want to add a walk-in patient to today's queue,
so that walk-ins and booked patients are in one list.
US-05 As a patient, I want to cancel my booking,
so that the slot can go to someone else and I am not counted as a no-show.The "so that" part is the most often skipped and the most useful. It tells the team what success means. If Aiko finds a simpler way to give patients the same benefit, the "so that" shows it would still meet the need.
The three Cs: card, conversation, confirmation
Ron Jeffries described a story as three parts in a 2001 article. The Card: "The card is a token representing the requirement"; it is short on purpose. The Conversation: "The requirement itself is communicated from customer to programmers through conversation". The Confirmation: the acceptance tests that show the story works. In other words, the card is a reminder to talk, not a complete specification.
Example:
CARD US-05 As a patient, I want to cancel my booking, so that the slot can go to
someone else and I am not counted as a no-show.
CONVERSATION (Nadia, Farhan, Kwame, 2026-10-15)
- Cancel online up to 2 hours before the visit; after that, phone the desk.
- Cancel from the link in the confirmation SMS, no password needed.
- A cancelled slot is free for others at once.
CONFIRMATION Given/When/Then scenarios from the previous page:
cancel well before, too late to cancel online, exactly two hours before.Notice that the confirmation reuses the acceptance criteria you wrote on the previous page. A story without confirmation is not ready to build, because nobody can say when it is done.
INVEST: checking a story
Bill Wake proposed the INVEST checklist in 2003: a good story is Independent, Negotiable, Valuable, Estimable, Small and Testable. It is a set of questions to ask, not a scoring system.
| Letter | Meaning | Weak story | Better |
|---|---|---|---|
| Independent | Can be built and released without waiting for another story | "Booking form, part 1" and "Booking form, part 2" | One story "Patient books a slot"; a second for "book for a family member" |
| Negotiable | Details are worked out in conversation, not fixed on the card | "…with a blue Book button, 40 px high, top right" | "…book a slot"; Omar and Nadia agree the design in conversation |
| Valuable | Delivers something a user or customer cares about | "Create the bookings database table" | "Patient books a slot", which includes the table it needs |
| Estimable | The team understands it well enough to size it | "Improve the reminders" | "Patient gets an SMS reminder at 18:00 the day before" |
| Small | Fits easily in an iteration | "Patient manages all their appointments online" | Split into book, cancel, reschedule, see my bookings |
| Testable | There is a clear way to confirm it works | "Booking should be user-friendly" | "4 of 5 patients over 60 book without help within 3 minutes" |
In Wake's article, "small" means at most a few person-weeks of work. Many teams today prefer much smaller stories, a few days each, so that several fit into one sprint and progress is visible. Notice also that "Create the bookings database table" is real work, but it is a task inside a story, not a story: on its own, no user can do anything new.
Splitting large stories
An epic is a story too big to build in one iteration. Epics are useful early, as placeholders for big areas of work, but they must be split before the team can plan them. The rule that matters most: split vertically, so that each small story still goes through every layer (screen, logic, data) and gives a user something working. Splitting horizontally, by layer, produces pieces that have no value until all of them are finished.
Example:
Horizontal split (avoid) Vertical split (better)
------------------------------- -------------------------------------------
Story A: all database tables US-01 Patient books a free slot
Story B: all API endpoints US-05 Patient cancels a booking
Story C: all screens US-09 Patient reschedules a booking
US-12 Patient sees their upcoming bookings
Nothing works until A, B and C Each one works on its own and can be shown
are all finished. to Nadia at the end of the iteration.Common ways to split an epic, with ClinicQueue examples:
| Split by | Epic: "Patients manage their appointments online" |
|---|---|
| Workflow step | Book, then cancel, then reschedule, then see upcoming bookings |
| Kind of user or data | Book for myself first; book for a family member later |
| Business rule | Cancel any time first; add the 2-hour rule as a separate story |
| Happy path, then exceptions | Book a free slot; then handle "slot just taken"; then "no free slots in 14 days" |
| Simple first, richer later | Reminder with fixed text; later the AI-drafted text checked by staff |
A good split lets the customer decide to stop after any piece. If Nadia says "cancel is enough, skip reschedule for now", the team has lost nothing.
Use cases
A use case describes how an actor (a person or another system) uses the system to reach a goal, step by step. It has a main flow (the normal path where everything goes well) and alternative flows (what happens at a given step when something is different or goes wrong). Use cases hold more detail than stories, and teams write them in many slightly different styles.
Example:
UC-01 Book a slot
Primary actor: Patient
Other actors: SMS company (sends the confirmation)
Goal: The patient has a confirmed booking with a doctor.
Precondition: The clinic has published its opening hours and doctors' slots.
Main flow
1. The patient opens the booking page and chooses a doctor and a day.
2. The system shows the free slots for that day.
3. The patient chooses a slot and enters a name and phone number.
4. The system checks the slot is still free and saves the booking.
5. The system shows a confirmation with date, time and doctor.
6. The system asks the SMS company to send a confirmation SMS.
Alternative flows
2a. No free slots that day: the system shows the next day with free slots.
4a. The slot was taken a moment ago: the system says so and shows the next three
free slots; the patient continues at step 3.
4b. The phone number is not valid: the system asks the patient to correct it.
6a. The SMS company does not accept the message: the booking stays confirmed; the system
retries once after 10 minutes and then lists the failure on the desk screen.
Postcondition: The slot is taken and appears in that day's queue.Writing alternative flow 6a forced a decision: if the SMS fails, is the booking still valid? Nadia said yes. That kind of question is exactly what a short story card leaves to the conversation, and what a use case makes the team answer in writing.
Story or use case?
| Situation | Better fit | Why |
|---|---|---|
| Planning and ordering work in short iterations | User story | Small, easy to estimate and reorder |
| Many branches, errors and other systems involved | Use case | Alternative flows make every branch visible |
| Customer available for frequent conversation | User story | Details can be agreed when needed |
| Written agreement needed (contract, regulator, distant team) | Use case | The detail is written down, not only spoken |
Many teams use both: stories in the backlog for planning, and a use case for the few interactions that are complex. ClinicQueue keeps stories for everything and writes a use case only for booking and reminders, where the SMS company makes failures likely.
Prioritising
There is always more to build than time to build it. Prioritising means deciding, with the customer, what comes first and what is left out. Two simple techniques cover most needs.
MoSCoW
MoSCoW is attributed to Dai Clegg (1994) and later became central to DSDM, an Agile method. Its four categories, as the Agile Business Consortium describes them:
- Must Have: without it the release is useless or unlawful. The Must Haves together are the "Minimum Usable SubseT (MUST)" the project guarantees.
- Should Have: important, but the release still works without it, perhaps with a workaround.
- Could Have: nice to have; the first to drop if time runs short.
- Won't Have this time: agreed to be out of this release. "This time" matters: it is a decision for now, not a rejection forever.
The Consortium's guidance is "typically no more than 60% Must Have effort" and "around 20% Could Have effort". This is guidance, not a law, but the reason is sound: if almost everything is a Must, there is no room to absorb a surprise, and any slip breaks the release. The Could Haves are the contingency.
Value versus effort
Inside each MoSCoW group, a team still has to choose an order. A simple tool is to give each story a value score (here 1 to 5, agreed with Nadia) and an effort estimate (here in story points, the relative sizes you will meet on Planning, Estimation and Scheduling), and place it on a grid.
Example:
low effort (1-3 points) high effort (5+ points)
+-----------------------------------+----------------------------------+
high value | QUICK WINS: do early | BIG BETS: plan carefully |
(4-5) | US-02 reminder, US-03 queue, | US-01 book a slot |
| US-04 walk-in, US-05 cancel | |
+-----------------------------------+----------------------------------+
lower value | FILL-INS: when there is room | AVOID FOR NOW |
(1-3) | US-06 arrived, US-07 language, | US-10 AI-drafted reminder, |
| US-08 doctor delay, US-09 resched.| US-11 online payment |
+-----------------------------------+----------------------------------+Value and effort are both estimates, so the grid is a conversation starter, not an answer. It also ignores dependencies: US-01 is a big bet, but nothing else works without bookings, so it goes first anyway.
A product backlog for ClinicQueue
A product backlog is the single ordered list of everything the product might need. In the 2020 Scrum Guide its commitment is the Product Goal, the target the backlog works towards. Lena runs a backlog session with Nadia, Omar, Farhan, Aiko and Kwame. Nadia decides value; the developers give sizes; together they agree the order. All values and sizes are invented.
Example:
Product Goal: patients book and are reminded online, so that phone calls and missed visits fall.
Order ID Story (short) MoSCoW Value Points
----- ----- ---------------------------------------------- ------- ----- ------
1 US-01 Patient books a free slot Must 5 5
2 US-03 Receptionist sees today's queue Must 5 3
3 US-04 Receptionist adds a walk-in to today's queue Must 4 2
4 US-02 Patient gets an SMS reminder the day before Must 5 3
5 US-05 Patient cancels a booking (2-hour rule) Should 4 2
6 US-06 Receptionist marks a patient as arrived Should 3 1
7 US-07 Patient chooses Bangla or English reminders Should 3 2
8 US-08 Doctor marks a delay; waiting patients get SMS Could 3 3
9 US-09 Patient reschedules a booking Could 2 2
-- US-10 AI drafts reminder text, staff approve it Won't 2 8
-- US-11 Patient pays online Won't 2 13Check it against the MoSCoW guidance, using points as a rough measure of effort for release 1 (the Won't Haves are not counted):
Must = 5 + 3 + 2 + 3 = 13 points
Should = 2 + 1 + 2 = 5 points
Could = 3 + 2 = 5 points
Total = 23 points
Must share = 13 / 23 = 56.5% (guidance: no more than about 60%)
Could share = 5 / 23 = 21.7% (guidance: around 20%)Why this order? Bookings come first because the queue, reminders and cancellations all need them. The queue screen and walk-ins come next because Rina cannot drop her paper diary until both work. Reminders follow, then the Should Haves in order of value for effort. US-10, the AI-drafted reminder, is "Won't have this time": it needs the safety check from the previous page (a person approves each text) and depends on reminders working first. It stays in the backlog for a later release.
The backlog is never finished. After every review, new stories arrive, some are split, and the order changes. The planning page starts from this backlog to break the work down and estimate it.
Where this shows up at work
- Issue trackers in most teams hold the backlog as stories with acceptance criteria and an order; learning to write good ones is a daily skill.
- Product owners and managers spend much of their time ordering the backlog and explaining why.
- Freelancers can use MoSCoW with a client to agree a fixed budget: Musts are promised, Coulds are delivered if time allows.
- Interviews for junior roles sometimes ask you to write or split a user story on the spot.
Common mistakes
- Technical tasks written as stories. "As a developer, I want a database table" delivers nothing to a user. Fix: put the task inside the story that needs it.
- Stories with no "so that". The team builds the action and misses the point. Fix: always write the benefit, and check the work against it.
- Splitting by layer. Nothing can be shown until every layer is done. Fix: split vertically by workflow step, rule or user type.
- Everything is a Must. There is no room for surprises and the priority list means nothing. Fix: keep Must effort well below the total and treat Coulds as contingency.
- Treating "Won't" as "never". Stakeholders fight every Won't because they think it is lost. Fix: say "Won't have this time" and keep the item in the backlog.
Try it yourself
- Easy: Check this story against INVEST and rewrite it: "As a user, I want the system to be fast and to have a nice booking page and reminders."
- Medium: Split the epic "As the receptionist, I want to run the day's queue, so that patients are seen in a fair order" into at least four vertical stories. Give each one acceptance criterion.
- Hard: Nadia's budget changes: release 1 can hold only 15 points. Using the backlog above, (a) choose what goes into release 1 and say why; (b) compute the Must share and say what the MoSCoW guidance warns about; (c) Nadia then asks to make US-10 (AI-drafted reminders) a Must. Write your reply in three or four sentences.
Answers
- Problems: "a user" is not a real role; "fast" is not testable; it is not small and is hard to estimate, because it mixes three things (speed, booking page, reminders); there is no "so that", so the value is unclear. A rewrite as three stories: "As a patient, I want to book a free slot online, so that I do not have to phone the clinic"; "As a patient, I want an SMS reminder the day before, so that I do not forget my visit"; and "fast" becomes a non-functional requirement or acceptance criterion: "the free-slots page loads within 3 seconds for 95% of requests". A good answer names at least three INVEST problems and turns the vague quality into a measurable criterion.
- Model answer: (1) "As the receptionist, I want to see today's booked patients in time order, so that I know who is next." Criterion: a booking made online appears in the list within 5 seconds. (2) "…add a walk-in patient to today's queue, so that everyone is in one list." Criterion: a walk-in added at 09:30 appears after the patient booked for 09:30. (3) "…mark a patient as arrived, so that the doctor sees who is waiting." Criterion: an arrived patient is shown as "waiting" on the doctor's view. (4) "…remove a patient who left without being seen, so that the queue stays accurate." Criterion: the removed patient disappears from the list and the next patient moves up. A good answer gives each story a role, an action and a benefit, keeps each one usable on its own, and gives a concrete, testable criterion.
- (a) The four Musts total 13 points, leaving 2. The next story in backlog order, US-05 cancel (2 points, value 4), fits exactly; US-06 (1 point, value 3) would leave 1 point that no remaining story fits. So release 1 = US-01, US-03, US-04, US-02 and US-05, 15 points. (b) Must share = 13 / 15 = 86.7%, far above the guidance of about 60%. With almost no Should or Could work to drop, any slip in a Must breaks the release. Options: more time or budget, a smaller Must (for example a simpler first queue screen), or accepting the risk openly. (c) Model reply: "US-10 is 8 points, so as a Must the release would need 21 points, more than the 15 we have. It also needs reminders working first and a person checking every text before it goes out. I suggest we keep it as Won't have this time for release 1 and plan it first for release 2." A good answer computes the numbers correctly, separates "what fits" from "what is ordered first", and explains the AI decision with dependency, capacity and safety reasons.
Summary
- A user story says who wants what and why ("As a … I want … so that …"); Ron Jeffries's three Cs describe it as a card, a conversation and a confirmation.
- Bill Wake's INVEST checklist (2003): Independent, Negotiable, Valuable, Estimable, Small, Testable.
- Split epics vertically, by workflow step, rule, user type or happy path first, so each piece works on its own.
- A use case lists an actor's main flow and alternative flows; it fits complex interactions and situations that need written detail.
- MoSCoW (attributed to Dai Clegg, 1994; central to DSDM) sorts work into Must, Should, Could and Won't have this time; keep Must effort to about 60% or less, and order within groups by value for effort and dependencies.
- The product backlog is one ordered list towards a Product Goal; it changes after every review.
Next: Project Management and the Roles in a Software Team looks at who does what on the ClinicQueue team, and how scope, time, cost and quality are balanced.

