Most UK buyers asking how much custom software costs need a usable range, not a vague “it depends”. At Code23, current public starting bands (reviewed 27 August 2026) are:

Discovery Sprint: £750 to £1,500
Custom SaaS MVP / Web App: £18,000 to £40,000
Custom Multi-Vendor Marketplace: £25,000 to £60,000
Custom AI & Agentic Integration: £30,000 to £100,000+

These are Code23 starting bands, not universal market averages. Final price follows scope, risk, integrations and what you need the first release to do in production. Full public project bands live on our pricing page.

What custom software is
When custom software helps, and when it does not
Current Code23 cost bands
What drives the cost
Pricing models and trade-offs
How to reduce cost without cutting the wrong things
What to give an agency for a credible estimate
Build cost versus ongoing cost
Frequently asked questions
Next step
References

What is custom software?

Custom software is built around your workflows, data and commercial rules. Off-the-shelf tools ship a fixed product with configuration limits. A bespoke system can model the way your team actually works: roles, approvals, pricing logic, inventory rules, partner onboarding or reporting that a generic product never quite fits.

That does not make custom the default answer. Many teams get further, faster, with a well-chosen SaaS product plus light integration. Custom work earns its place when the product is the business advantage, when licence fees and workarounds start to dominate, or when you need to own the roadmap.

When custom software helps, and when off-the-shelf is better

Custom software is useful when you need:

Workflows that mirror how your staff, customers or partners actually operate
Integrations and data models that packaged tools only half support
A product you can extend without waiting on a vendor backlog
Clear ownership of IP, hosting choices and release cadence

Stay with off-the-shelf when the job is standard (CRM contact tracking, basic accounting, simple marketing site CMS) and the packaged tool already covers 80% of the need at a predictable subscription. Buying software you can configure is often cheaper than building software you then have to maintain.

Code23 builds custom SaaS platforms, websites and apps, multi-vendor marketplaces and AI systems. We have operated under the Code23 brand since 2005. For a concrete product example, see the eSHP case study.

Current Code23 cost bands

Buyers often ask about a custom CRM, a customer portal or “an app like X”. The honest answer is still a band, because two CRMs can differ by roles, migrations, reporting depth and the number of systems they must talk to. Use the public Code23 bands below as a planning frame, then refine through discovery.

Project type	Starting band	Typical timeline	What it usually covers
Discovery Sprint	£750 to £1,500	1 to 2 weeks	Interactive prototype, architecture, schema and a locked fixed-price build band (credited toward the build)
Custom SaaS MVP / Web App	£18,000 to £40,000	6 to 10 weeks	Auth, roles, billing hooks, dashboards, APIs, automated tests and warranty
Custom Multi-Vendor Marketplace	£25,000 to £60,000	8 to 12 weeks	Multi-party payments, onboarding, buyer flows and payouts
Custom AI & Agentic Integration	£30,000 to £100,000+	6 to 20 weeks	Retrieval, workflow agents, evaluation harnesses and operational controls

Bands include the design and engineering needed to ship the agreed first release. They do not automatically include every third-party licence, paid API usage, multi-year support retainer, or later product expansion. Scope, risk and operational requirements decide where you land inside (or above) a band. Compare the full set on Code23 pricing.

What drives the cost of custom software

Price is mostly a function of uncertainty and surface area. The more you clarify early, the tighter a fixed price can be. The drivers below are the ones that move estimates in practice.

Uncertainty and discovery

If nobody has mapped the primary user journeys, the estimate is guessing. A short discovery sprint turns fuzzy ambition into screens, data shapes and a build band you can actually buy. Skipping discovery often looks cheaper, then reappears as change requests mid-build.

Workflows, roles and business rules

Every distinct role, approval path and exception rule adds screens, permissions and test cases. A three-role admin tool is not the same product as a marketplace with buyers, sellers and operators. Complexity is not vanity features; it is the logic that must stay correct under real use.

Integrations and payments

Connecting accounting, CRM, ERP, shipping or payment providers can dominate the schedule. Some APIs are clean; others need mapping work, webhooks, retry behaviour and reconciliation. Payments add compliance and edge cases. Sequence non-critical integrations after the first release when you can.

Data migration

Moving existing data into a new system is often under-scoped. Dirty records, duplicate customers, inconsistent statuses and partial history all take engineering time. Clean what you can before migration. Move only the data the new product needs for day one.

Security and compliance

Auth, audit trails, least-privilege access, encryption in transit and at rest, and evidence for customer security questionnaires all add work. Treat security as a build requirement, not a polish pass. The OWASP Application Security Verification Standard is a practical checklist for what “secure enough” can mean at different assurance levels.

Platforms and expected scale

A responsive web app is usually cheaper than parallel native iOS and Android clients. User volume matters too: expected concurrency, file storage, search load and reporting all shape infrastructure choices. Build for the users you have evidence for, with a clear path to scale, rather than imagining every future channel on day one.

UX, accessibility and design depth

Clear information architecture, usable forms and accessible components protect adoption. Heavy illustration, motion systems or multi-brand theming raise cost. Accessibility is not optional decoration; WCAG 2.2 sets the shared bar for perceivable, operable, understandable and robust interfaces.

Testing, infrastructure and release quality

Automated tests, staging environments, CI/CD, monitoring and rollback plans cost money up front and save money later. Cutting them to hit a headline price is how quiet defects become expensive incidents.

Deadline pressure

A fixed launch date with unfinished decisions forces parallel work, overtime or scope cuts under stress. Soft deadlines let you sequence risk. Hard deadlines are fine when the commercial reason is real; budget for the compression.

Pricing models and trade-offs

Three commercial shapes show up most often:

Fixed price after discovery. You pay for a short discovery, then buy a locked build price for an agreed scope. Best when the first release can be defined. Change outside that scope is re-priced, which protects both sides.
Time and materials. Flexible when the problem is still shifting. You need strong backlog discipline and regular demos, or cost drifts without a clear product outcome.
Dedicated product team / continuous delivery. A retained capacity model for roadmap work after launch. Useful when you will keep shipping every month; less useful if you only need one contained MVP.

No model always wins. Fixed prices reward clear scope. Time and materials reward learning speed when scope cannot settle yet. Retained teams reward continuous product ownership. Match the commercial shape to how settled the problem is.

How to reduce cost without cutting the parts that protect the product

Cost control is mostly sequencing and decision speed:

Narrow the first release. Ship the workflow that creates value. Park secondary roles and reports.
Validate workflows before you build them. GOV.UK’s service manual is clear that user research improves service design; even lightweight observation of real tasks beats guessing screens.
Clean the data early. Migration cost tracks mess more than row count.
Sequence integrations. Launch with the one or two systems that unblock revenue or operations.
Reuse proven components. Auth, billing, email, file storage and admin patterns rarely need reinvention.
Make decisions quickly. Waiting a week for a button label still burns calendar time. Agile delivery works when teams keep a short feedback loop; see GOV.UK’s core agile principles.

Do not “save” money by deleting testing, access control, backups or accessibility. Those cuts usually return as incidents, rebuilds or blocked enterprise deals.

What to give an agency for a credible estimate

A useful briefing pack includes:

Who the primary users are and what success looks like in the first 90 days
The must-have workflows for release one, written as steps rather than feature wish lists
Systems that must integrate on day one versus later
Rough data volumes and whether a migration is required
Compliance or security expectations from customers or insurers
Platform constraints (web only, or native apps too)
A real deadline with the commercial reason attached
Budget range you can actually approve, even if it is a band

If that pack is thin, start with discovery rather than forcing a fake fixed price. Vague briefs produce either padded quotes or optimistic quotes that break mid-project.

Build cost versus ongoing cost

The build invoice is only part of the total. Plan for:

Hosting and environments: production, staging, backups and CDN
Third-party tools: payments, email, SMS, search, analytics, error tracking
Monitoring and security maintenance: dependency updates, uptime checks, incident response
Support and iteration: bug fixes after warranty, small improvements, seasonal changes

Ongoing cost should match how actively the product will change. A stable internal tool needs less continuous delivery than a marketplace competing every week. Code23 publishes current care and growth options on the pricing page; treat those as living figures rather than copying old package prices from archived blog posts.

Frequently asked questions

How much does it cost to develop custom software?

At Code23, planning bands start at £750 to £1,500 for discovery and £18,000 to £40,000 for a custom SaaS MVP or web app, with marketplaces and AI builds higher. Exact cost follows scope. See current pricing.

How much does custom software development cost in 2026?

Use current public bands, not archived 2020 figures or third-party dollar averages. Code23’s bands above were reviewed on 27 August 2026 and remain starting ranges, not market averages.

How much does it cost to develop a custom app?

A focused web app MVP often lands in the £18,000 to £40,000 band at Code23. Native mobile clients, heavy offline behaviour or multi-sided marketplaces push the band up. Start with the platform your users already live on.

How much does software development cost?

“Software development” covers everything from a brochure site to a regulated platform. Ask for a band against a named first release, then fund discovery if the release is still fuzzy.

How do I estimate the cost of software development?

List users, release-one workflows, integrations, data migration and deadline pressure. Map that to a published band, then run a short discovery to lock a fixed build price. Estimates without those inputs are theatre.

Have you got a custom software project in mind?

If you already know the problem and the first release, bring the briefing pack above and we can place you in a realistic band. If the problem is still forming, start with a Discovery Sprint so you buy clarity before you buy a build.

Product design and discovery helps shape the journeys; pricing shows the current bands; contact Code23 when you want a concrete next step.

References

GOV.UK Service Manual: How user research improves service design
GOV.UK Service Manual: Core principles of agile
OWASP: Application Security Verification Standard
W3C: Web Content Accessibility Guidelines (WCAG) 2.2

How Much Does Custom Software Cost?

Contents