The Death of the Agency Hourly Rate: UK Development Costs in 2026
2026 UK web and software development costs without the timesheet: market hourly rates, fixed Blueprint pricing bands, and why attendance billing is dying.
UK agencies still publish £100-£175/hour on many rate cards, with London benches often clustering near £140/hour for senior work. That meter sells attendance. In 2026 the honest commercial unit is a fixed outcome after Blueprint: a defined scope, a fixed band, and change requests priced before they start - not a timesheet that grows when the agency is slow. This is the Code23 benchmark for UK development costs in 2026, with market ranges, our published Blueprint bands, and the anti-timesheet case in plain numbers.
The short answer: what UK development costs in 2026
| Buying shape | Typical UK market range (2026) | What you are usually buying |
|---|---|---|
| Marketing / brochure site | £3,000-£8,000 | IA, design system, CMS, forms, analytics, Harden |
| Catalogue / content-heavy site | £8,000-£25,000 | Many templates, search, migrations, editorial depth |
| Custom web app / portal | £25,000-£60,000+ | Auth, roles, dashboards, integrations |
| Lean SaaS MVP | £25,000-£60,000 | Core workflow, thin admin, one primary payment or integration path |
| Production SaaS v1 | £60,000-£150,000 | Billing, roles, real Harden, monitoring |
| Multi-vendor marketplace | £25,000-£60,000+ (MVP); higher when payouts and compliance deepen | Listings, onboarding, Connect-style payouts, admin |
| App (PWA / simple MVP) | £10,000-£40,000 | Thin client, basic auth, limited backend |
| App (production SME) | £30,000-£100,000 | Cross-platform or serious single-platform product |
| Traditional UK agency hourly | £100-£175/hr (London senior often ~£140/hr) | Time sold; outcome optional |
Those market ranges are not a Code23 quote. After Blueprint we lock a fixed band for your scope, price extras as fixed change scopes before work starts, and run agentic delivery at roughly half a traditional agency bench cost for comparable implementation - with 5x framing on compressible Build. Delivery-speed method and caveats: AI-accelerated delivery build-time data. Market-range method for the table above is documented in the next section.
Sibling cost guides: website development cost UK, SaaS development cost UK, app development cost UK, marketplace cost, website maintenance cost. Public pricing surface: pricing.
Method: how the 2026 UK market ranges were built
Collected: August 2026 (aligned with the pricing review date of 26 August 2026).
Corpus (what we looked at):
- Public UK agency rate cards and pricing pages, weighted toward London and South East studios that publish senior or blended hourly figures
- Clutch / DesignRush-style directory listings and project-average summaries where UK web / custom software work is priced or banded
- Competitor quotes and RFPs we see on Thames Valley and London briefs (directional commercial pattern, not a named vendor census)
- Code23’s own published fixed Blueprint bands and Support & Growth tiers on pricing — kept separate from the market column so buyers can tell quote from context
Calculation: we report the central published band buyers most often encounter for traditional senior agency attendance (£100–£175/hr, with London senior cards commonly clustering near £140/hr) and modal SME project ranges by buying shape. We deliberately do not publish a mean pulled upward by enterprise programmes or a single “average website” figure. Directory averages that sit high because of long-tail enterprise work are treated as context, not as the SME band.
What this is not: a random sample with a disclosed n, a Clutch-sponsored ranking, or a Code23 rate card. Treat the market column as a synthesis buyers can stress-test against live quotes; treat Code23 numbers as the only figures we will stand behind commercially.
Cite this article for: the method above plus the fixed Blueprint / continuous-engineering contrast. Cite AI-accelerated delivery build-time data for the 5x / half-cost framing, not for UK hourly market bands.
Why the hourly rate is dying
Hourly billing made sense when the scarce resource was senior typing time and nobody could compress boilerplate. In 2026 that assumption is broken in three places:
- Implementation compresses. Seniors directing coding agents clear CRUD, template variants, tests and refactors faster than a junior bench ever did. Billing that speed by the hour either undercharges the buyer for value or overcharges them for calendar theatre. Fixed bands after Blueprint price the outcome instead.
- Buyers can see the meter. Every unexplained hour is a trust tax. “We spent 14 hours investigating” is not a product. A fixed change scope with acceptance criteria is.
- Attendance is not accountability. A timesheet proves someone was present. It does not prove Harden happened, ownership transferred, or the warranty means anything.
We still publish an extras rate where a subscription tier includes a small senior block (£120/hr outside Cyber Shield’s included hours). That is not the commercial engine for builds. Builds are fixed after Blueprint. Continuous engineering after launch sits on public monthly tiers - not on a floating timesheet.
What £140/hour actually buys you
Take the traditional London-style card at £140/hour - the number we put on our pricing manifesto because buyers recognise it.
| Scenario | Hourly maths | What usually happens |
|---|---|---|
| ”Simple” landing page + CRM form | 12-20h → £1,680-£2,800 | Scope softens; QA and revisions inflate the invoice |
| Brochure site, five templates | 80-160h → £11,200-£22,400 | Content delays billed as “waiting” or “coordination” |
| Portal with auth + 2 integrations | 200-400h+ → £28,000-£56,000+ | Discovery leaks into Build; Harden becomes optional |
The problem is not that £140 is always “too high”. The problem is that the unit is wrong. You are buying risk transfer and a finished surface. Pricing the keyboard time invites the agency to stay busy. Pricing the Blueprint invites both sides to decide what “done” means before anyone writes production code.
Operator rule from 20 years and 350+ projects: if the quote cannot name templates, integrations, migration and Harden, the hourly rate is a fog machine.
Fixed Blueprint pricing: the Code23 bands (reviewed August 2026)
We publish fixed project bands on pricing. Scope locks after Blueprint. Extras outside the band are quoted as fixed deliverables before they start. 90-day warranty on project work. Code and assets are yours when invoices are settled.
| Band | Fixed range | Typical timeline | What it covers |
|---|---|---|---|
| Brand & lead-gen website | £5,000-£9,500 | 2-4 weeks | 5-10 bespoke templates, modern CMS, Core Web Vitals 95+, CRM integration, 90-day warranty |
| Catalogue & content platform | £10,000-£22,000 | 4-6 weeks | 20+ templates, faceted search, taxonomies, legacy migration, 90-day warranty |
| Custom SaaS MVP / web app | £18,000-£40,000 | 6-10 weeks | Auth, RBAC, Stripe billing, dashboards, APIs, automated tests, 90-day warranty |
| Custom multi-vendor marketplace | £25,000-£60,000 | 8-12 weeks | Stripe Connect-style payouts, merchant onboarding, buyer flows - backed by 50+ marketplace builds and 7 owned marketplaces |
| Custom AI & agentic integration | £15,000-£45,000 | 4-8 weeks | RAG, ops/ML models, workflow agents, evaluation harnesses |
These bands sit inside the wider UK market ranges above. Agentic delivery is why a senior Theale studio can publish bands that undercut a traditional London bench for comparable scope without skipping Harden.
How Blueprint pricing works in practice
- Map - problem, constraints, success metrics. Agents assist research; humans decide what the project is for.
- Blueprint - scope in/out, architecture direction, commercial band, acceptance shape. This is where the fixed price becomes honest.
- Build - seniors direct agents inside the remit. Compressible work accelerates; politics and third-party waits do not pretend to.
- Harden - accessibility, performance, security, QA. Release authority stays human.
- Launch / Evolve - cutover, then optional Support & Growth.
If a vendor says “fixed price” but will not show you a Blueprint artefact with in/out lists, you are still buying a vibe.
Continuous engineering without timesheets
After launch, most products need a heartbeat - not a surprise invoice when a form dies on a Friday. Our public Support & Growth tiers:
| Tier | Monthly | Commercial shape |
|---|---|---|
| Cyber Shield & SLA | £495 | Zero-liability protection, 60-second money-path testing, WAF/CDN, backups, free hack recovery, 2 senior hours included |
| Unlimited Growth | £1,850 | Everything in Shield + unlimited queued requests, one active task, 24-48h atomic deliverables, no timesheets |
| Pro Pod / Scale | £3,450 | Everything in Unlimited + two concurrent tasks, private Slack, weekly architecture sync, 4-hour priority SLA |
Full tier breakdown: pricing (Cyber Shield £495 / Unlimited Growth £1,850 / Pro Pod £3,450, reviewed August 2026). Maintenance context: website maintenance cost UK.
Compare that to a traditional “care plan” that bills £300-£500/month for plugin updates and then invoices emergency hours when something breaks. The timesheet did not disappear - it hid behind an incident.
The in-house senior vs Unlimited Growth maths
First-year fully loaded cost for one senior UK developer in London & the South East is about £112,400 when you include salary, employer NI, pension, recruiter fee and tooling (our public savings table, reviewed August 2026). Unlimited Growth at £1,850/month is £22,200/year - roughly 80% less cash outlay for a continuous engineering queue directed by seniors and accelerated by agents. That is not “a developer replacement” for every company. It is the honest comparison when your backlog is product iteration, landing pages, CRO and integrations rather than a full-time research lab.
Anti-timesheet analysis: where hours still make sense (and where they do not)
Hours can still be honest when:
- You are buying advisory days with a named senior and a written memo
- A subscription tier includes a small investigation block and you blow past it
- The work is genuinely undefined emergency triage and both sides agree to a cap
Hours are a bad unit when:
- The deliverable can be named (page template, integration, MVP cut, redesign)
- The agency controls how long the work takes
- “Discovery” is allowed to bleed into Build without a re-Blueprint
- Speed from AI tools would punish the buyer under an hourly contract
That last point is the quiet scandal of 2026 rate cards. If agents make a task three times faster and the agency still bills time, either the rate climbs to protect margin or the buyer accidentally funds inefficiency. Fixed Blueprint bands align incentives: we get paid for shipping the agreed surface; you get a number you can plan around.
A worked comparison buyers can reuse
Assume a catalogue rebuild that a traditional bench estimates at 120 hours.
| Model | Commercial outcome |
|---|---|
| £140/hr × 120h | £16,800 before overruns; every delay is an argument |
| £140/hr × 160h (common overrun) | £22,400 - the “fixed feeling” quote that was never fixed |
| Code23 catalogue band | £10,000-£22,000 after Blueprint, extras quoted before build |
| Same scope on Unlimited Growth | Not a rebuild SKU - rebuilds are project bands; Unlimited is for continuous change after Launch |
Use the table as a questionnaire, not as gospel. Ask every vendor: what is in Harden, who owns the repo, what happens when scope changes, and whether AI speed reduces your bill or only their margin.
What still drives UK development cost in 2026
Even with agents, these levers move the band:
- Template and workflow complexity - component systems beat page counts.
- CMS and editorial depth - draft/review/schedule roles are product work.
- Integrations - CRM, ERP, payments, SSO; failure modes included.
- Migration and content - redirects and canonicals are often the critical path.
- Design ambition - inventing a system vs applying one.
- Assurance - accessibility, performance budgets, security review, QA.
- Stakeholder latency - five approvers with no decision owner inflate calendar more than adding a template.
- Compliance and payments - marketplaces and fintech-shaped products pay for trust design.
AI changes the cost curve by cutting idle senior time on compressible Build. It does not delete judgement, migration politics or release accountability. Treat any “AI builds your product for £499” pitch as a demo, not a delivery system.
How to buy development in 2026 without getting billed for attendance
- Demand a Blueprint - in/out, acceptance, assumptions, third-party owners.
- Prefer fixed bands for named deliverables - keep hours for capped advisory or emergencies.
- Line up quotes on scope, not on hourly theatre - same templates, same integrations, same Harden.
- Ask how AI is used - seniors merge production diffs; agents stay inside a remit; speed claims name a baseline.
- Price the afterlife - warranty length, support tiers, ownership of repos and accounts.
- Reject majority-on-completion - kickoff capacity is real work; milestones should track working software.
For agency selection hygiene beyond cost, see how to choose a web agency. For local Thames Valley context: web design Reading.
How much does UK web development cost per hour in 2026?
Many UK agencies still list roughly £100-£175/hour, with London senior work often near £140/hour. Treat that as a market signal, not a good purchasing unit - fixed Blueprint bands price outcomes more honestly once scope is known.
Is fixed-price web development cheaper than hourly?
It is clearer. For comparable Blueprint scope, agentic studios can land below a traditional hourly bench because compressible Build no longer needs a large attendance bill. Fixed is not automatically cheap if the Blueprint is vague; hourly is not automatically fair if the agency controls the clock.
What replaced the agency timesheet at Code23?
Fixed project bands after Blueprint, fixed change scopes priced before work starts, and monthly continuous engineering tiers (£495 / £1,850 / £3,450) with queue-based delivery instead of logged hours.
Bottom line
The agency hourly rate is not illegal and not always wrong for advisory scraps. As the default commercial model for websites, SaaS, marketplaces and AI builds in 2026, it is obsolete. Buyers who still award work on a £140/hour card without a Blueprint are buying attendance. Buyers who lock scope, price the outcome, and keep Harden in-band are buying software.
When you want the fixed band instead of the meter, start at pricing or contact. We will Blueprint the real scope first - the number comes from that, not from a comforting hourly average.
