Outlook
Part of Web design agencies outlook: UK evidence limits
Three scenarios for England's web agencies and what would disprove each
Three scenarios for England's web agencies, each with an assumption, leading indicators, disconfirming evidence and an action owner at a December 2026 cut-off.
England's web-agency market has no published growth rate. The ONS business registers count enterprises by SIC code, and no single code isolates website design sold to clients. The answer to whether demand is rising or falling is that the reviewed evidence does not settle it.
Three bounded scenarios follow. Each carries an assumption, leading indicators, disconfirming evidence and an action with a named owner. No probabilities are assigned, and the evidence cut-off is December 2026.
What to take away
- No growth rate for England's web-agency market exists in the ONS sources reviewed, because web work sits inside SIC 62012, 62090 and 73110 alongside software development and advertising.
- Three scenarios follow, each paired with the evidence that would disprove it, and none carries a probability.
- Watch named England sourcesFind a Tender award notices, the GDS accessibility monitoring report and Companies House filings.
- Recheck every scenario at the December 2026 evidence cut and before any 2027 commitment.
Why the ONS registers cannot price this market
The ONS UK business activity, size and location dataset breaks down by region, industry and employment size band. Web design work is spread across SIC 62012, 62090 and 73110. None of those rows isolates agency demand, project value or delivery capacity.
Business demography counts VAT or PAYE registered enterprises and reports births, deaths and survival. It does not report order books, day rates or pipeline value. Both releases are annual, so confirm the reference date shown on the dataset page before quoting a figure.
Two agencies in one SIC row rarely match in scope, so run benchmark comparability checks before comparing anything.
Three scenarios and what would disprove each
Core assumption
- A
- Briefs demand accessibility, privacy, security and exit evidence before appointment
- B
- Buyers remediate or extend an existing site instead of rebuilding
- C
- AI drafts research, content, code or tests while human assurance stays
Evidence that would disprove it
- A
- Comparable briefs in your sample omit those gates
- B
- A defined cohort funds full replacements after like-for-like appraisal
- C
- A supplier sample reports no such use, or shows no acceptable benefit
Scenario A: buyers require deeper assurance
Assumption: more England briefs make accessibility, privacy, security, provenance and exit evidence conditions of appointment. The GOV.UK accessibility monitoring report is the closest published England-wide test method, covering public sector sites and mobile apps.
Leading indicators: recurring clauses in your own tender sample, requests for accessibility test evidence, software component records, incident routes and data export demonstrations.
Scenario A: assurance gates
Leading indicators
- Tender clauses
- Recurring gates
- Accessibility
- Tests requested
- Component records
- Software records
- Incident routes
- Routes shown
- Export demos
- Demonstrations
Disconfirming evidence
- Tender clauses
- Gates omitted
- Accessibility
- Narrower boundary
- Component records
- Not required
- Incident routes
- Accepted without
- Export demos
- No exit evidence
Disconfirming evidence: comparable briefs omit those gates, or buyers accept a narrower service boundary. Action: the procurement owner updates the evidence schedule, not a market size estimate.
Example: a council or NHS brief published through Find a Tender normally names an accessibility statement, a conformance target and a named owner.
Scenario B: repair competes with replacement
Assumption: some organisations remediate or extend an existing site rather than fund a rebuild. No official source measures how often England buyers choose that route.
Example: a retailer whose checkout fails after a platform change may buy an audit and a fix list rather than a new build. Find a Tender notices show whether public bodies buy audits and support packages or full replacements.
Disconfirming evidence: a defined buyer cohort authorises complete replacements after a like-for-like appraisal. Action: the service owner compares acceptance, defects, security exposure, migration risk, whole-life cost and exit.
Agreeing that comparison in writing is easier with the commercial contracts for web design agencies checklist, which lists the records that keep a commission honest.
Scenario C: assisted production increases review work
Assumption: AI drafts research, content, code or test material while human assurance stays. The ONS AI in business analysis comes from a voluntary survey and cannot establish agency adoption or productivity.
Leading indicators: suppliers disclose model-dependent tasks, buyers request provenance and evaluation records, and release notes log AI-related corrections. The NCSC secure AI guidance supplies lifecycle questions, not certification.
Disconfirming evidence: a defined supplier sample reports no such use, or a controlled comparison shows no acceptable benefit. Action: the delivery owner runs the assisted and existing routes on the same authorised fixture and logs every correction.
Before you act
- Define the eligible opportunity set and the collection period.
- Name the evidence owner before reading any signal.
- Keep the scenario log beside the decision it informed.
- Drop any indicator that never changes an action.
- Recheck in December 2026 and before any 2027 commitment.
A clean company record proves little about delivery quality, as the England agencies with verified company records review shows.
Common questions
Is England's web-agency market growing or shrinking?
The reviewed ONS releases do not isolate it, so no growth rate can be quoted. Business demography reports enterprise survival across broad industries, not agency order books.
What would disprove Scenario A?
Comparable briefs omitting accessibility, privacy, security and exit gates, or buyers accepting a narrower service boundary. The procurement owner then updates the evidence schedule at the next live procurement.
When should the scenarios be rechecked?
At the December 2026 evidence cut, and immediately before any 2027 decision. A fresh source cut may support a different set of scenarios.
