1. Double-counting
The same benefit is counted under two different social value topics. Example: a rent reduction counted once as reduced financial stress and again as improved life satisfaction.
How It Works
Hundreds of Topics. One Connected System.
Many social-value approaches assess outcomes separately and combine them later. ONEMODEL instead represents connected outcomes within one shared system, so attribution, interactions and boundaries can be examined together.
SVT's claim is narrower and stronger: it measures hundreds of topics together, person by person, in one connected model. That joined-up approach is where the difference lies. This is not just positioning - it follows from how ONEMODEL is built, and it changes which errors the model can make.
ONEMODEL measures hundreds of topics together in one connected model, person by person. The distinction is architectural: outcomes share the same people, boundaries, time and counterfactual.
How Common Approaches Differ
| Approach | How It Works | Common Limitation to Manage |
|---|---|---|
| WELLBY methods | Take survey data, compare to population baselines, monetise using a fixed $/WELLBY conversion | Assumes all people's satisfaction points are equivalent; one topic at a time; no person-level paths |
| SROI | Assign monetary proxy values to selected outcomes, often using external evidence or value banks | Context transfer, counterfactual design and overlap between outcomes depend on how the individual study is constructed |
| HACT Social Value Bank | Apply published wellbeing values to selected outcomes | The value-bank approach does not itself provide a connected person-level system across all modelled outcomes |
| Hedonic pricing | Infer the value of attributes from observed market-price differences | Observed prices capture capitalised preferences but do not themselves represent complete person-level welfare pathways |
| Bespoke multi-outcome studies | Combine several outcome estimates in a tailored analytical framework | Cross-outcome coherence, overlap control and reproducibility depend on how the separate estimates are integrated |
| Multi-model analysis | Use separate specialist models for population, housing, fiscal, health or other domains | Definitions, boundaries and flows must be explicitly reconciled across modules before results can be treated as one system |
These approaches can produce useful insights. The key difference is that their architecture does not necessarily enforce one shared population, set of boundaries and counterfactual across every outcome. ONEMODEL does.
Six Risks to Control in Fragmented Analysis
When outcomes are estimated in separate analytical components and combined later, six risks require explicit control:
2. Missing mechanisms
An effect is stated as a result but has no modelled pathway. Example: a report claims a housing programme "reduces loneliness" - but there is no represented mechanism connecting housing to social connection.
3. Category drift
Definitions shift between modules. Example: one team's "household" includes single-person dwellings; another's does not. Their outputs cannot be combined without creating a fictional aggregate.
4. Conservation violations
People, money, or time appear or disappear between analytical domains. Example: the labour model creates jobs that the fiscal model does not fund.
5. Frame dependence
Conclusions change when the same situation is described from a different angle. Example: a programme looks positive when measured from the participant's perspective but negative from the community's - with no reconciliation.
6. Open boundaries
Flows leave one model but never arrive in another. Example: health savings from a housing programme are claimed in the housing evaluation but never appear in the health budget - because the two models do not share boundaries.
How ONEMODEL Controls These Risks
ONEMODEL is a single-system architecture. Hundreds of social-value topics are generated from the same model, person by person. That design is intended to prevent, expose or reconcile these risks within one accountable analytical structure:
| Failure Mode | How ONEMODEL Prevents It |
|---|---|
| Double-counting | Benefits tracked within one system - a gain counted under one topic cannot also appear under another |
| Missing mechanisms | Every change arises from a represented process - nothing is inserted by assumption |
| Category drift | One set of definitions, units, and classifications used throughout |
| Conservation violations | People, money, and time are conserved across the entire model - nothing appears or disappears |
| Frame dependence | Conclusions remain stable regardless of which perspective examines them |
| Open boundaries | All boundary flows are closed or explicitly externalised - nothing leaks between topics |
This is not achieved through post-hoc checking. It is achieved through architectural design. The errors cannot occur because the system structure does not permit them.
ONEMODEL satisfies the S4CI integrity framework. In plain terms, S4CI is a seven-part check that asks whether the model describes a world that could actually exist:
The S4CI Standard
- 1. Single-system structure
- Is this one accountable architecture - or separate models stapled together?
- 2. Typed consistency
- Do all parts use the same definitions, units, categories, and time steps?
- 3. Causality
- Do changes come from represented mechanisms - or are results inserted by assumption?
- 4. Conservation
- Are people, money, and time tracked without duplication or leakage?
- 5. Invariance
- Do conclusions survive when you look at the same situation from a different angle?
- 6. Closure
- Are all boundaries complete - does everything that leaves one place arrive somewhere?
- 7. Coherence
- Does the whole compose into one intelligible system?
A model that passes all seven is decision-grade: its outputs describe a situation that could actually happen and can be defended in real decisions.
What This Means for You
If you commission SVT, your evidence will:
- Never contain hidden double-counting - because all social value topics are tracked in one system
- Always trace to a mechanism - every result can be explained through a represented causal path
- Show distribution - who gains, who loses, by how much, through what channels
- Survive scrutiny - the same conclusion holds regardless of which reviewer examines it from which angle
- State its assumptions - every parameter is declared and sensitivity-tested
You will not receive a number pulled from a database. You will receive evidence generated from a model of the actual situation - specific to your population, your context, and your intervention.
Common Questions
Can double-counting occur in other social value methods?
Yes. When social value topics are valued separately and added together, the same benefit can be counted under more than one topic. ONEMODEL prevents this by generating all topics within one system - a benefit counted once cannot appear again.
What does "decision-grade" mean?
A model is decision-grade when its outputs would be defensible as the basis for a real decision. It is higher than "plausible numbers" (which can be produced by defective models) but does not claim certainty.
Is SVT saying all other methods are wrong?
No. Other methods can produce useful insights. SVT identifies a specific structural limitation - fragmented architecture - that permits certain failure modes. A single-system design prevents those modes. This is a statement about architecture, not about the competence of other practitioners.
How can I verify these claims?
SVT's methodology is documented at journal-publication standard. The S4CI framework is published as an independent research contribution. Both are available for independent review and scrutiny.
Ready for Evidence That Holds Together Under Challenge?
Schedule a consultation.
Start a ConversationExternally reviewed methodology, developed over 20+ years400+ engagements (CANCEA, 2002-2026)Featured in the ICPM Social Infrastructure Blueprint (2026)