Asking "what will this software cost?" before defining the scope is like asking a builder for the price of a building without saying how many floors it has. The answer is either a guess or a sales number. A useful budget is built from scope, uncertainty and the cost of running the system after launch.
This guide gives you a cost model, a calculator structure you can put in a spreadsheet and the questions that expose an incomplete proposal. It does not publish a fixed price, because a credible price depends on the work.
Why a single number is a bad plan
A custom software project has four kinds of cost, and most quotes cover only the second:
- Discovery: understanding the process, users, data and constraints.
- Delivery: design, build, test and launch the first version.
- Change: the work that appears when real users touch the system.
- Run: hosting, support, upgrades and internal ownership after launch.
A quote that covers delivery but not discovery is an estimate built on assumptions. A quote that covers launch but not run is a first-year price, not a budget.
The cost model
Build the estimate from lines, each with an owner and a range.
- Discovery and requirements: interviews, process maps, scope document, acceptance criteria. Price it as a defined phase.
- Product and UX design: screens, states, permissions and mobile behaviour. More screens and roles mean more design work.
- Engineering: person-weeks by role, multiplied by a blended rate. Senior time costs more but usually reduces rework.
- Quality assurance: test cases, automation, regression and user acceptance support. Do not treat this as optional.
- Project management and business analysis: coordination, decisions, change control and reporting.
- Integrations: each external system has its own access process, sandbox, limits, error handling and approval. Count them individually.
- Data migration: extraction, cleanup, mapping, import, reconciliation and rollback.
- Infrastructure and licences: hosting, database, storage, email, SMS, maps, payment or AI usage.
- Security and compliance: authentication, permissions, logging, backups, privacy review and testing.
- Training and documentation: admin guide, user guide, handover sessions and a support model.
- Contingency: a percentage agreed with finance, with a written statement of what it covers.
- Post-launch support: a monthly retainer or an internal budget for fixes and small changes.
A calculator you can build in a spreadsheet
Use a three-point estimate for each delivery line: optimistic, likely and pessimistic. The weighted estimate is:
weighted = (optimistic + (4 x likely) + pessimistic) / 6
Then calculate the project total:
total = discovery + (delivery person-weeks x blended rate) + fixed costs + contingency
And the monthly run cost:
run = hosting + licences + support + internal admin time
Keep the inputs visible. A spreadsheet with a single total hides the assumptions that matter: number of roles, integrations, migrated records and the confidence of the estimate. When a vendor gives you a number, ask them to fill in the same lines. If they cannot, the comparison is not like for like.
Contingency and change control
Contingency is not a secret buffer. It is the budget for things you know are uncertain: an API that behaves differently from its documentation, data that is dirtier than expected, or a policy decision that arrives late.
- Agree the percentage with finance and state what it covers.
- Keep a change log with the request, impact on scope, cost and date, and the approver.
- Review changes monthly, not at the end of the project.
- Separate defect fixes from new scope. A defect is included; a new rule is a change.
- Set a threshold above which a change needs owner approval.
Projects rarely fail because of one large change. They fail because twenty small ones were absorbed silently and the original date became fiction.
Whole-life cost
Plan for three years, not one. Include:
- Hosting and infrastructure growth.
- Licence and usage increases.
- Security patching and dependency upgrades.
- Support and small enhancements.
- Internal product ownership: someone must prioritise, test and approve.
- Training for new staff and branches.
- Exit or migration costs if the vendor relationship ends.
A system that is cheap to build but expensive to change is not cheap. The internal owner time is often the largest hidden line, especially in a company where the same manager also runs operations.
Questions to ask a vendor
- What is included in discovery, and what will we have at the end of it?
- Which roles, integrations and reports are in the fixed scope?
- What is explicitly excluded?
- Who owns the code, designs and data?
- How do you test, and can we see the test evidence?
- What happens when a third-party API changes or is unavailable?
- How do you handle data migration and reconciliation?
- What is the change request process and who approves it?
- What does support cost after the warranty period?
- Can we export our data and continue with another team?
A 90-day planning cycle
- Weeks 1-2: map the process, users, exceptions and data sources.
- Weeks 3-4: write the scope and acceptance criteria; count roles and integrations.
- Weeks 5-6: request estimates on the same line items and compare assumptions.
- Weeks 7-8: check references, security practice and support model; agree the contract shape.
- Weeks 9-10: run a paid discovery or a narrow pilot if the risk is high.
- Weeks 11-12: approve the budget, contingency and run cost; set the first review date.
Metrics for the project
- Estimate accuracy: actual cost divided by the approved estimate.
- Change cost as a share of the original budget.
- Defects found after launch, by severity.
- Cycle time from approved change to release.
- Adoption: active users against the intended user group.
- Uptime and support ticket volume after launch.
Where this fits
If you need a scoped build rather than an open-ended hourly arrangement, start with Custom Software & Integrations. The pricing page explains how discovery, delivery and support are separated. For the annual budget view, read how to plan an SME IT budget.
Sources
- GOV.UK Service Manual, "How the discovery phase works", accessed 17 November 2026.
- CISA, "Secure by Design", accessed 17 November 2026.
- ISO/IEC 25010:2023, Systems and software quality models, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.