52,481 people. Skeina makes visible what happens between services — who was unseen, what they needed, who accepted responsibility, how long it took, whether anything was delivered, and whether it helped.
All figures on this page are synthetic demonstration data — no real people
Not another dashboard of everything at once. One funnel, from found to a person-verified outcome — and one failure state the neighbourhood must see. Click any stage.
173 of 302 confirmed needs reached a person-verified outcome: 113 resolved, 48 improved, 12 accepted endpoint — because housing, frailty and caring often cannot truthfully be called “resolved”.
See the loop board →4 service gap · 3 eligibility gap · 5 capacity gap · 2 boundary gap. Cannot be administratively closed; escalates on a clock. This tells commissioners exactly what intervention is required.
See the ownerless board →The dangerous space is not the wait — it is the time when nobody owns what happens next. Skeina measures it.
See zero-owner days →Who are we responsible for — and who are we not seeing? Coverage is the honest denominator. Previously unobserved means not previously visible within the relevant neighbourhood monitoring pathway, discovered through another legitimate point of contact.
Signals from GP records, hospital, pharmacy, family hubs, care, schools, community — and residents themselves. Published deterministic conditional rules: every consequential output can be traced to its data, rule, evidence source, version and accountable human decision. Current release: no statistical inference or language model.
Who accepts the next responsibility — by name, on a clock? Ownership is Skeina’s distinctive proposition. Until someone accepts, the need is OWNERLESS, and that is an outcome the neighbourhood must see.
Did something actually happen? Acceptance and delivery are measured separately — a referral accepted then ignored is a different failure from one nobody takes. Routed to real services by their Wolverhampton Information Network identifier.
Did it help, against the outcome the person themselves sought? Checked again at 30 and 90 days; reopened needs are counted, not forgotten. “Referral sent” is not a success measure anywhere in Skeina.
Where did the system fail — ownerless needs, capacity gaps, inequalities, silent misses, rules that need to change? The register is versioned, silence itself is audited, and every learning has somewhere explicit to land.
A fall reported by home care, and a medicines contradiction between two records.
2 of 3 signals confirmed by a human. One sent back — and the send-back is kept.
Yes. Lena decides — declining would be recorded and respected, not chased.
“I want to stay living at home.” Skeina doesn’t just ask whether she wants help — it knows what success means.
GP and pharmacist, by name. Until someone accepts, the need is OWNERLESS — and escalates.
Face-to-face medication review; strength-and-balance referral by WIN identifier.
Against her own outcome, Lena says: improved.
38 days in total. Checked again at 30 days: sustained — reopened needs are counted, not forgotten.
Two published rules fired: SK-INT-014 (fall in the last 90 days → strength and
balance, home hazard check) and SK-INT-050 (sources contradict → surface both
versions to a human; never auto-resolve). Each rule shows its condition, its authority, its
named owner and who holds the decision. The engine proposes; people decide.
Open the live model to click through 125 journeys like this →
Most dashboards report what their algorithm did. Skeina deliberately reports what it may have missed, what nobody owns, and where help ran out.
Confirmed needs no organisation has accepted. Cannot be administratively closed; escalate to neighbourhood governance on a clock.
See the ownerless board →Where no rule fired, a random sample goes to independent review — 1 missed actionable need found this cycle (estimated miss rate 5.0% in this synthetic sample). Reviewed for whether a rule should exist. Its mirror is also measured: non-actionable findings among those surfaced.
See the silent-case lab →How many times people told the same story before their need reached an outcome. Housing 4 · mental health 3 · family hub 1. Target: once.
See the experience measures →“No wrong door” is usually rhetoric. Skeina measures it: does reaching an outcome depend on whether you first appeared at the GP, the pharmacy, the school gate or the Jobcentre?
See equity by entry route →Where case-finding met a wall. Reported as commissioning intelligence — latent demand revealed, not programme failure.
See need vs capacity →Every rule has an owner, a source, a version and a review date. Deterministic conditional logic — no statistical inference, no language model.
View the register with performance cards →Nothing left the neighbourhood until a named person reviewed them.
See why →Findings a professional sends back, modifies or escalates. No “healthy range” is prescribed — Skeina watches for abnormal patterns instead: 100% acceptance, zero challenge, one site that never sends anything back.
How decisions are held →Same structured inputs, same register version, same outputs — every time, demonstrated in the live model. Reproducibility is not clinical validity; those are tested separately.
Run it yourself →Where a domestic-abuse indicator is present, SK-SAFE-001 blocks
every outbound referral until a named professional completes a safe-contact check — because a
well-meaning letter home can be dangerous. Holds are visible, owned, and stay in the statistics
after release.
A community interest company with an asset lock — and proposed resident governance rights: two resident directors, a young-person seat, a carer seat, and a resident challenge process for any expansion of data use. Provider-held is not the same as neighbourhood-owned, so both are designed in.
Read the full proposal →One care home, thirty retrospective records: rule debugging, workflow and data-quality testing with PDSA cycles — explicitly not outcome evaluation. Phase 1 is prospective operational feasibility (100–300 people); Phase 2 is neighbourhood evaluation with a comparator.
The staged plan →has precedent somewhere — Frome, Findhelp, Foundry, Kinzigtal, Bradford. The integration of those components into one neighbourhood accountability protocol is our hypothesis — Skeina should be tested comparatively, not accepted on assertion.
The international landscape →Agree the measures before we measure anything; an independent evaluator as a condition of support. The commissioner receives disclosure-controlled aggregate returns — person-level information needed for coordination remains within the governed care environment and is never exposed through the commissioning return.
The population health case →