Connect to Work: Employer Capability Is the Delivery Test

Connect to Work is a delivery challenge, not a referral exercise

Connect to Work arrives at a moment when the UK labour market remains marked by persistent inactivity and uneven routes back into work. The economic inactivity rate for people aged 16 to 64 was 20.9% in April to June 2026, while the employment rate was 75.1%. The Office for National Statistics advises against over-reading short-term movements and notes that labour-market measures should be interpreted together because their sources and coverage differ (Office for National Statistics, 2026) ↗.

That caution matters. The inactivity rate is a useful statement of the national challenge, not a measure of Connect to Work’s target population and not a performance baseline from which local commissioners can infer a programme effect. Connect to Work is aimed at people who are disabled, have health conditions or face other barriers to employment, including some people at risk of leaving work. It is voluntary, locally commissioned and designed around supported-employment approaches commonly described as ‘place, train and maintain’ (Department for Work and Pensions, 2026) ↗.

The programme’s significance lies in the commissioning responsibility it creates. Accountable bodies must translate national expectations into a local service model, delivery route and performance approach. Delivery-plan guidance sets out staged funding arrangements extending to 31 March 2030 (Department for Work and Pensions, 2026) ↗. That is long enough for places to learn and improve, but only if they treat implementation as more than building referral pathways between health, employment and skills services.

**Sanctuary interpretation.** The decisive test will be whether local systems alter the practical conditions in which people get, keep and progress in work. A referral can identify need; it cannot itself make a vacancy suitable, an induction accessible or a line manager confident enough to resolve difficulties early. Connect to Work should therefore be understood as a local labour-market intervention with an employer-facing operating model, not simply as an additional employment-support service.

Supported employment has evidence behind it—but no automatic result

The national guidance requires accountable bodies to commission or establish provision using IPS and/or the Supported Employment Quality Framework, with associated fidelity requirements and a focus on competitive employment and sustained outcomes. It also warns that payment-by-results arrangements may not work well for the supported-employment model (Department for Work and Pensions, 2026) ↗. This is an important design signal: the intended model depends on continuing relationships and problem-solving, not on a quick hand-off once a job start has been recorded.

The wider evidence supports that direction while setting limits on the claims that can reasonably be made. An overview review and meta-analysis of IPS beyond severe mental health found higher job-entry rates than control services. Across its large and medium studies, the estimated odds ratio was 1.32. However, the authors also identified substantial heterogeneity, incomplete fidelity, small samples and inconsistent outcome measurement (Whitworth et al., 2024) ↗. An odds ratio is not a promise of a fixed percentage-point gain in every locality, sector or participant group.

**What follows from the evidence.** Fidelity matters because it protects the active logic of supported employment: a preference-led search for suitable work, active employer engagement, appropriate job matching and continuing support after appointment. But fidelity should not be confused with standardising every local interface. Places can adapt access routes, partnerships and sector focus to local conditions; they should not dilute the model into generic employability provision that ends at a vacancy referral.

There is a further trade-off. A narrow focus on rapid job entry can improve apparent throughput while placing people into roles that have poor fit, unstable hours or inadequate support. Conversely, excessive caution can slow job search and restrict participant choice. The task for commissioners is not to choose between speed and quality in the abstract. It is to build a service capable of moving quickly towards roles that are viable for both participant and employer, then intervening promptly when that viability is at risk.

Employer capability is the missing infrastructure

‘Place, train and maintain’ is often described as though it were a linear participant journey: assess an individual, find a vacancy, arrange training and measure whether employment is sustained. In practice, sustained work is co-produced. It depends on the participant and employment specialist, but also on how a job is specified, how work is allocated, whether recruitment and induction are accessible, and whether a manager can respond when a problem emerges.

**Sanctuary interpretation.** This makes employer engagement a delivery function rather than a communications workstream. A partnership board can secure employer logos without changing any workplace practice. The more useful unit of change is frequently the line manager, the immediate team and the role itself. A manager may need support to distinguish essential from customary job requirements, redistribute a limited set of tasks, establish predictable check-ins, understand adjustment processes and know where to seek help without breaching confidentiality.

For SMEs, this distinction is especially material. A small business may not have a dedicated human-resources function or capacity to navigate a lengthy multi-agency process for a single role. It may, however, value a named contact who can help clarify a vacancy, support a suitable match and resolve issues after appointment. That is not a case for treating employers as passive recipients of subsidy. It is a case for reducing the practical transaction costs of inclusive recruitment and retention.

An employer operating model should therefore include segmentation of local employers; a clear offer to managers; vacancy and task analysis; support for job carving or redesign where appropriate; fast escalation when placements are at risk; and systematic learning from failed matches. This capability can also matter for local productivity. Avoidable turnover, repeated recruitment and unmanaged absence impose costs on employers, while sustainable employment can strengthen household income and local spending. Those local-value effects are plausible rather than guaranteed, and should be tested through local data rather than assumed.

The UK trial evidence makes local learning non-negotiable

The Health-led Employment Trials provide a useful corrective to both scepticism and complacency. The evaluation used random allocation to test time-limited IPS in primary-care settings for people with moderate to mild mental and physical health conditions. At 12 months, the West Midlands out-of-work group had a statistically significant four-percentage-point increase in the proportion achieving at least 13 weeks of employment: 22% in the treatment group compared with 18% in the control group. In Sheffield City Region, the out-of-work group had no statistically significant employment impact, despite positive health and wellbeing effects. The in-work group showed a smaller, lower-confidence employment effect (Department for Work and Pensions, 2023) ↗.

The evaluation found limited evidence of an overall effect on earnings or job-quality indicators. Yet its qualitative findings suggested that personalised support, job carving and job design could improve how participants experienced job quality (Department for Work and Pensions, 2023) ↗. Taken together, these findings support the model’s potential while showing why a programme name alone cannot explain outcomes.

**Sanctuary interpretation.** Local labour demand, participant mix, relationships between health and employment services, employer access and implementation quality are all credible routes through which results may differ. The trial cannot identify a single universal cause of the difference between areas, so commissioners should resist simplistic league-table thinking. Equally, variation is not an excuse to abandon measurement. It is a reason to ask sharper questions: Which sectors produce sustainable roles? Where do placements break down? Are some groups receiving lower-quality opportunities? Which features of employer support appear to prevent exits?

A credible local evaluation should connect these questions to delivery decisions. It should not merely report starts and sustained outcomes upward after the fact. Commissioners need feedback quickly enough to redesign practice, provider support and employer offers during delivery.

A commissioning agenda that links participant support to workplace change

The national guidance already places weight on employer engagement, fidelity assurance, performance management and local evaluation (Department for Work and Pensions, 2026) ↗. The opportunity is to connect these requirements into one operating system rather than assign each to a separate workstream.

**Sanctuary recommendations.**

1. **Define the employer offer before scaling referrals.** Set out what an employer receives: help with vacancy and task analysis, candidate matching, manager briefing, adjustment navigation, retention troubleshooting and a named contact. Test the offer with SMEs as well as larger anchor employers.

2. **Measure job quality alongside sustained employment.** Job starts and 13-week outcomes remain important, but they are incomplete. Local dashboards should also examine contracted hours, predictability of work, participant-reported fit, retention at later milestones, progression and reasons for exit. The Health-led Trials underline why employment alone is not a sufficient account of value (Department for Work and Pensions, 2023) ↗.

3. **Use fidelity review as improvement, not compliance theatre.** Require routine assessment against the selected supported-employment model, publish actions arising from review and protect core features when adapting locally. The IPS evidence indicates that uneven fidelity and outcome measurement complicate interpretation of results (Whitworth et al., 2024) ↗.

4. **Turn placement data into employer and system intelligence.** Aggregate recurring reasons for declined offers and failed placements: travel, inflexible hours, inaccessible recruitment, task design, induction or absence procedures. Use those findings to improve provider practice, business-support offers and the local skills response.

5. **Equip line managers as delivery partners.** Provide short, practical support on work allocation, check-ins, adjustments, confidentiality boundaries and escalation routes. Broad awareness activity has a place, but it is unlikely to substitute for help at the point where a manager has to make work workable.

6. **Commission for learning as well as volume.** Compare outcomes across participant groups, sectors and delivery routes. Where differences emerge, investigate whether they reflect access, job quality, local demand or implementation. Funding and contract management should give providers enough stability to act on that learning, consistent with the guidance’s caution on payment-by-results models (Department for Work and Pensions, 2026) ↗.

The lasting outcome should be more capable local employers

Connect to Work should be judged by whether people enter and remain in suitable work. But its longer-term value may depend on a second outcome: whether participating employers become better able to recruit, manage and retain people whose skills are too often missed by conventional hiring processes.

That does not mean shifting responsibility from public services to employers, or subordinating participant preference to business convenience. Supported employment works only when the employment goal is meaningful to the individual and support continues when a job begins. Nor does it mean that every employment barrier can be solved through job redesign; local demand, health, transport and financial circumstances continue to matter.

It does mean that local leaders should make workplace co-production visible in governance, commissioning and measurement. By the end of the current delivery period in March 2030, the strongest places will not simply be able to report referrals and starts. They will be able to show what they learned about creating sustainable roles, which employer practices changed, and how that learning improved opportunities for subsequent participants (Department for Work and Pensions, 2026) ↗.