
Foundations
Part of Web design agencies foundations: evidence and rules for 2027
Four web agency delivery models compared on one buyer need
Compare four web-agency delivery models in England using one buyer need, common commercial fields, evidence gaps and independent review gates.
What to take away
- Compare four web agency delivery models on one buyer need, one period and common obligations, not headline fees.
- In England, Torchbox and Fat Beehive sell fixed projects, while Cyber-Duck and Substrakt sell continuing services.
- Equal Experts and Made Tech supply embedded specialists; Brainlabs and Journey Further sell outcome-linked work.
- Each model sells a different unit, so scope change, capacity and acceptance terms must be defined.
- Fixed projects strain when discovery reveals unknowns; variation and stop authority matter more than launch dates.
- Outcome-linked models need agreed measures and attribution limits, with legal and security gates still applying.
- Rights, data roles and security evidence must be settled before handover, not left until the end.
Model fit in one line: fixed for known scope, continuing for steady demand, embedded for buyer-led teams, outcome-linked for measurable targets.
The common buying case
Assume an organisation in England needs discovery, content migration and accessible templates. It also needs development, two integrations, launch assurance and a year of support. The comparison period is twelve months from appointment. Money is in pounds sterling, and VAT treatment stays blank until a proposal is verified.
Those are comparison units, not a recommended scope or forecast. This comparison is model-level desk research dated 6 September 2026. The agencies named below are public examples of a model, not recommendations or rankings.
No single model wins. A buyer with a clear scope and acceptance point should start with a fixed project. A buyer with steady demand should test a continuing service. A buyer who leads delivery should test embedded specialists. A buyer with a measurable target should test outcome-linked work.
Four delivery models compared
Fixed project
- Unit purchased
- accepted deliverables
- Scope change
- priced variation
- Capacity evidence
- delivery plan
- Acceptance
- artefact and test
- Exit
- files and rights
Continuing service
- Unit purchased
- queue or capacity
- Scope change
- priority rule
- Capacity evidence
- monthly commitment
- Acceptance
- work-item criteria
- Exit
- backlog and exports
Embedded specialists
- Unit purchased
- named roles and time
- Scope change
- buyer backlog
- Capacity evidence
- availability by role
- Acceptance
- buyer approval
- Exit
- knowledge transfer
Outcome-linked
- Unit purchased
- outcome boundary
- Scope change
- dependency review
- Capacity evidence
- resourcing vs outcome
- Acceptance
- metric plus evidence
- Exit
- data and duties
Fixed project: Torchbox, Fat Beehive
- Unit purchased
- accepted deliverables
- Scope change
- priced variation
- Capacity evidence
- delivery plan
- Acceptance
- artefact and test record
- Price and VAT (typical, England, 2026)
- fixed fee per phase, commonly £25,000 to £90,000 excluding VAT for a build of this size
- Exit
- files, rights and credentials
Continuing service: Cyber-Duck, Substrakt
- Unit purchased
- defined queue or capacity per month
- Scope change
- priority rule
- Capacity evidence
- monthly commitment
- Acceptance
- work-item criteria
- Price and VAT (typical, England, 2026)
- monthly fee, typically £750 to £4,000 excluding VAT, set by capacity and response times
- Exit
- backlog, exports and access removal
Embedded specialists: Equal Experts, Made Tech
- Unit purchased
- named roles and time
- Scope change
- buyer backlog
- Capacity evidence
- availability by role
- Acceptance
- buyer approval by increment
- Price and VAT (typical, England, 2026)
- day rate, typically £400 to £1,000 excluding VAT by role, seniority and clearance
- Exit
- knowledge transfer and access closure
Outcome-linked: Brainlabs, Journey Further
- Unit purchased
- agreed outcome boundary
- Scope change
- dependency review
- Capacity evidence
- resourcing against outcome
- Acceptance
- metric plus service evidence
- Price and VAT (typical, England, 2026)
- base fee plus a performance element, often 5% to 15% of the agreed measure
- Exit
- data, calculation and unresolved duties
Which model fits when. Fixed project fits when scope, acceptance and content are clear. Continuing service fits when demand is steady and a queue can be defined. Embedded specialists fit when the buyer leads delivery and can manage roles. Outcome-linked work fits when a measure and attribution rule can be agreed.
Worked cost illustration for the common buying case. Applying the typical ranges above over twelve months:
- £25,000 to £90,000excluding VATFixed project
- £750 to £4,000per monthContinuing service
- £400Embedded specialists
- 5%Outcome-linked
These ranges are desk-research illustrations, not quotes. Check them against a written proposal that sets out the same scope, period and obligations.
Four named models and where each strains
Fixed project: Torchbox and Fat Beehive. Torchbox, based in Charlbury, Oxfordshire, built the open-source Wagtail CMS, which charities and public bodies use. Fat Beehive, in London, builds websites for charities and publishes budget guidance for the sector.
A fixed project gives a clear acceptance point when content and integrations are understood. It strains when discovery reveals major unknowns, so the variation route and stop authority outweigh a promised launch date.
Continuing service: Cyber-Duck and Substrakt. Cyber-Duck, in Elstree, Hertfordshire, sells design, build and ongoing support to financial services and public sector clients. Substrakt, in Birmingham, works with arts and culture organisations.
The unit is a defined queue or capacity per month. The contract must define the queue, clock and priority. It must also fix unused capacity, renewal and emergency work. Support without those fields is not a comparable service unit.
Embedded specialists: Equal Experts and Made Tech. Equal Experts, headquartered in London, supplies teams to UK government departments and large enterprises. Made Tech, also in London, is listed on AIM and delivers digital services for the public sector.
Embedded teams work inside the buyer's process. That can cut handovers, and it raises access and role ambiguity. GOV.UK employment-status guidance says tax and employment-law status can differ and depends on the actual relationship. A freelancer label does not decide it.
Outcome-linked: Brainlabs and Journey Further. Brainlabs, in London, and Journey Further, in Leeds, are performance media agencies where fees can follow measured results. The model ties payment or continuation to a measure.
It needs an agreed population, numerator, denominator and period. The data source, baseline, exclusions and treatment of outside changes must be settled too. Attribution is not proof that the agency caused the result. Legal, accessibility, security and privacy gates still apply even if the metric improves.
Non-compensating evidence gates
Rights cannot be left until handover. The IPO's commissioned-work guidance says the creator is generally the first copyright owner unless otherwise agreed in writing. List pre-existing code, third-party components, content, design assets and new work, with an ownership or licence outcome for each.
Evidence gates before appointment
- List pre-existing code, content and design assets
- Review data roles using ICO guidance
- Assign named security responsibilities
- Ask every bidder for the same schedule
- Mark absent answers as unknown
Data roles follow actual decisions. The ICO controller and processor guidance should inform a qualified review of analytics, forms, hosting and support access. A contract title alone cannot allocate statutory status.
Security evidence needs named responsibilities. The NCSC online-service guidance covers access, secure development and testing. It also covers logging, incident handling and supplier management. A cheaper fee or a stronger portfolio does not offset missing assurance.
Ask every bidder to return the same schedule: currency, VAT treatment, period, included quantity and exclusions. Add assumptions, expenses, licences, the change unit and acceptance tests. Include the support boundary and exit, then mark absent answers unknown.
The best model is the one whose verified obligations fit the buyer's service, not the one with the simplest label.
Before you act
- Define one buyer need and a twelve-month comparison period.
- Agree rights and licence outcomes for all work and assets.
- Review data roles and security responsibilities with qualified advice.
- Ask every bidder to return the same schedule of obligations.
- Mark absent answers as unknown rather than assuming.
Common questions
Why is comparing headline fees across delivery models meaningless?
Because each model sells a different unit: accepted deliverables, monthly capacity, named roles and time, or an agreed outcome boundary. Without a common buyer need, period and obligations, price comparisons lack a fair basis.
What must a continuing service contract define to be comparable?
It needs a defined queue, a priority rule and a monthly capacity figure. The buyer should also see how unused capacity, renewal and emergency work are handled.
How should a buyer handle rights and data roles before handover?
Record ownership or licence outcomes for each asset and take qualified advice on data roles. The ICO controller and processor guidance is a starting point.
Which agencies in England sell each delivery model?
Fixed projects: Torchbox in Charlbury, Oxfordshire, and Fat Beehive in London. Continuing services: Cyber-Duck in Elstree, Hertfordshire, and Substrakt in Birmingham. Embedded specialists: Equal Experts and Made Tech, both in London. Outcome-linked work: Brainlabs in London and Journey Further in Leeds.



