Skip to content

Scope and minimum viable product

Status
Approved
Version
1.0.0
Authority
Recommended practice
Applies to
Core
Markdown source

Summary

Scope is the line around what a release will and will not do. Writing it down, with an explicit out-of-scope list, lets everyone say no to good ideas without a quarrel and lets changes happen in the open. A minimum viable product (MVP) is the first release defined by what it must teach the team: the smallest complete, usable version that tests the riskiest assumption about the users. Scope and MVP boundaries are decisions of the person who owns the product, not of the people who build it.

Rules

  • R1. Every product and every release MUST have a written scope statement that lists what is in scope and what is out of scope. The out-of-scope list MUST name the notable things a reader might otherwise expect.
  • R2. The scope statement MUST be kept in the requirements specification (or linked from it) and MUST separate what is deferred to a later release from what is a non-goal of the product.
  • R3. Scope, the MVP boundary and priorities MUST be decided by the person who owns the product. Anyone else, including an AI agent, MUST NOT widen or narrow scope on their own; they MUST raise the proposal as a question or change request.
  • R4. An item that is out of scope MUST NOT be built until the scope has been changed through R7.
  • R5. An MVP MUST be defined by a stated hypothesis: who the users are, what the team believes about them and which evidence will confirm or refute it. The definition MUST name the scenarios (by id) that are in the MVP, the ones that are not, and the measure and date by which the MVP will be judged.
  • R6. An MVP MUST deliver at least one complete path from the user’s start to the user’s result, with acceptance criteria that pass for it. A set of unfinished parts that cannot be used end to end MUST NOT be called an MVP.
  • R7. A change of scope MUST be recorded as a change request that states what changes, why, and its effect on other requirements, schedule and cost; it MUST be approved by the person who owns the product, and the specification, the roadmap and the affected issues MUST be updated in the same or a linked change.
  • R8. When the judging date of an MVP is reached, its result MUST be compared with the stated measure and the decision (continue, change direction or stop) MUST be recorded.
  • R9. In-scope requirements SHOULD be prioritised for each release on an explicit scale (for example Must, Should, Could, Won’t have this time), and the Must items SHOULD together be enough for the release to meet its goal.
  • R10. A request that arrives during the work and is not in scope SHOULD be recorded in the backlog with a note of who asked and why, rather than being built or silently dropped.
  • R11. The constraints that bound scope (budget, date, platform, team) SHOULD be recorded next to it.

Guidance

  • Where scope comes from. The vision’s outcomes and non-goals give the outer boundary; the scenarios and requirements fill it in; the scope statement fixes what this release takes. The IREB glossary defines scope as the range of things that can be shaped and designed when developing a system; Wiegers and Beatty split a project’s scope into what the first release includes, what later releases may include, and the limitations and exclusions.
  • Deferred or never. “Not in this release” and “not in this product” are different statements (R2). Keep both lists. The second is the vision’s non-goals; the first is the roadmap’s later milestones.
  • What an MVP is for. Ries defines the minimum viable product as the version that lets a team collect the most validated learning about customers with the least effort, and warns that it is not simply a minimal product. The Agile Alliance glossary repeats the definition. The test of an MVP is the question it answers, not the number of features it has.
  • Skateboard, not a wheel. Kniberg’s picture: a car built wheel by wheel gives the customer nothing until the end, while a skateboard, then a scooter, then a bicycle each work from the start. Choose the smallest thing that works end to end (R6), then grow it.
  • Prioritising. MoSCoW sorts items into Must, Should, Could and Won’t have this time. It works when the Must list is honest: if everything is Must, nothing is prioritised. Won’t items are written down so that they are not forgotten or re-argued.
  • Working with agents. Give an agent the scope statement and the out-of-scope list with the task. When the agent finds something useful outside scope, it records it as a question for the owner (R3, R10) and goes on with the task as scoped.
  • Small products. A short list of “in” and “out” in the specification is enough. Do not write a scope document for a bug fix.
  • When not to apply. Throw-away experiments do not need a scope statement; once their result is used by a product, the product does.

Explanation

Scope creep is rarely one big decision. It is many small agreements, each reasonable, none recorded, and the sum is a release that is late and unfocused. A written list of what is out, with an owner who alone can change it, turns each of those agreements into a visible, priced decision.

The MVP idea applies the same discipline to learning. A product is a bundle of guesses about users; the first release is the cheapest experiment that can test the riskiest guess. Ries’s point is that the aim is learning, not shipping less; Kniberg’s is that the thing shipped must still work end to end, or the learning is spoiled by users who cannot get through it. R5 to R8 put these two points into checks: a hypothesis, a complete path, a measure and a date.

Examples

Good — scope statement with a deferred list, non-goals and an MVP hypothesis:

## Scope of release 1 (MVP)
Hypothesis: engineers will use an offline notebook for every site visit if capture takes under 10 seconds.
Measure: 20 engineers use it on 80% of visits by 2027-03-31 (judged on that date).
In: S-A Capture a note offline (FR-1…FR-4), S-B Sync when online (FR-5, FR-6).
Deferred to release 2: S-C Photos in notes; S-D Search.
Non-goals of the product: shared editing; invoicing.

Bad — no hypothesis, no boundary, no date:

MVP: the basic version with the main features, to be extended as needed.

Sources