A custom software project does not need to begin with a hundred-page specification. It does need a shared foundation that clarifies the problem, supports comparison, and exposes uncertainties that could change scope, cost, or timing.
This checklist prepares an initial conversation or discovery stage. It is not a fixed specification and does not replace legal, tax, employment, privacy, accessibility, or security review appropriate to the context.
Before starting: define the decision you need to make
Do not assume that building software is the only answer. Preparation may lead to simplifying a process, configuring an existing tool, integrating systems, running a prototype, or deciding that the project is not yet viable.
The GOV.UK Service Manual recommends understanding the problem before committing to a build: users and goals, constraints, opportunities, the current cost of the problem, and evidence for deciding whether to proceed. Stopping is valid when the evidence does not justify the next stage.
Source: GOV.UK Service Manual — How the discovery phase works
1. Problem, outcome, and boundaries
- Describe the observable problem without embedding a solution. Prefer ‘onboarding repeats data across three systems’ to ‘we need a platform’.
- Explain who experiences the problem, how often, and what happens if it remains unsolved.
- Define the outcome to improve and how it connects to an organisational objective.
- State what is outside this stage so that scope does not expand by interpretation.
- Separate facts, assumptions, and open questions. Record the source of important data.
2. Users and the current journey
- Identify direct and indirect users: customers, operations, support, administration, security, and teams that will integrate or maintain the system.
- Map the current journey end to end, including email, spreadsheets, calls, approvals, and exceptions.
- Bring available evidence: interviews, observation, analytics, incidents, timings, or anonymised examples.
- Include accessibility needs, different levels of digital skill, and relevant alternative channels.
The GOV.UK Technology Code of Practice guidance notes that technology choices also affect support, operations, and future teams that will reuse or integrate what is built. Users are not limited to the people who see the interface.
Source: GOV.UK — Define user needs
3. Success and baseline
- Choose three to five signals that show whether the outcome works: completion, time, errors, rework, satisfaction, adoption, or operating cost, as appropriate.
- Record a baseline and data source; without one, improvement is difficult to distinguish from perception.
- Define frequency, owner, and limits for each metric. Do not turn a proxy into the product goal.
- Combine quantitative data with user research and operational signals.
GOV.UK recommends designing metrics with clear meaning and planning collection during the build, not afterwards. It also advises combining analytics with sources such as feedback, support, performance, and operating costs.
Source: GOV.UK Service Manual — How to set performance metrics
4. Scope of the first stage
- Define the priority flow and its start and end. An entire project is not a verifiable unit.
- List essential, desirable, and explicitly deferred capabilities.
- Write observable acceptance criteria, not only feature names.
- Identify the largest uncertainty and design a test to reduce it before expanding scope.
- Decide what would allow the team to proceed, correct, stop, or roll back.
5. Data, integrations, and migration
- For each data set, identify the source of truth, owner, sensitivity, quality, retention, and access.
- List external systems, owners, documentation, limits, test environments, and availability expectations.
- Clarify whether historical data must be imported, cleaned, reconciled, or kept in temporary coexistence.
- Provide current and peak volumes with their source; avoid projections without documented assumptions.
- Prepare synthetic or anonymised data where possible, and do not share secrets or unnecessary personal data during commercial discussions.
6. Security, privacy, and accessibility
- Classify the impact of errors, unauthorised access, downtime, and data loss.
- Define identity, permissions, logging, encryption, backup, recovery, vulnerability, and incident-response requirements proportionate to risk.
- Assign security and privacy review owners; confirm applicable obligations with qualified professionals.
- Include accessibility in goals, budget, responsibilities, design, content, and testing from the start.
NIST presents the SSDF as practices that can be integrated into different development lifecycles: prepare the organisation, protect software, produce well-secured software, and respond to vulnerabilities. Use it to make responsibilities and evidence concrete, not as an automatic guarantee.
Source: NIST — Secure Software Development Framework 1.1
W3C WAI recommends setting accessibility objectives, assigning responsibilities, allocating resources, and evaluating early and repeatedly. Considering it only at the end increases the risk that early decisions become expensive to correct.
Source: W3C WAI — Planning and Managing Web Accessibility
7. Operations and life after launch
- Decide who will deploy, monitor, respond to alerts, support users, and approve changes.
- Set proportionate operational objectives for availability, response, recovery, and log retention.
- Include dependency maintenance, patches, backups, infrastructure, licences, and support costs.
- Ensure the organisation can access repositories, environments, documentation, managed credentials, and recovery procedures.
- Plan knowledge transfer and an orderly exit; contractual terms require specific review.
8. People, decisions, and ways of working
- Name a product owner with authority to prioritise and accept outcomes.
- Identify specialists and owners who should participate: users, operations, data, security, privacy, accessibility, finance, and legal where appropriate.
- Agree channels, demonstration cadence, decision records, change management, and escalation.
- Clarify the real availability of key people and dependencies on other teams.
9. Budget, timing, and assumptions
- State budget and dates as known constraints or ranges, not as a technical estimate that has already been validated.
- Separate discovery, build, migration, stabilisation, operation, support, and third-party costs.
- Document what each figure includes, its assumptions and uncertainty, and what could change it.
- Allow for research, testing, security, accessibility, data, documentation, and adoption—not only coding.
What to bring to the first meeting
- One page covering the problem, users, outcome, boundaries, and main unknowns.
- A simple map of the current process and systems.
- The available baseline and three success signals.
- Known data, integrations, constraints, and owners.
- Budget or range, relevant date, and why those constraints exist.
- The decision you expect to make at the end of the first stage.
You do not need every answer. Useful preparation makes questions visible and orders them by risk. The best next step is the smallest one that produces enough evidence for a responsible decision.
What an MVP should include to validate a B2B product
Codiloop can help turn this checklist into a bounded discovery and a verifiable first stage.

