evidoria

How the quality score is computed

Every practice in Evidoria carries a single quality score from 0 to 100. It is calculated the same way in every catalogue, so scores are comparable and explainable.

The method, in one line

score = 100 × the (weighted) average of each evaluation dimension, after rescaling every dimension to 0–1.

  1. Each catalogue has a rubric: a set of evaluation dimensions (e.g. transferability, impact, sustainability).
  2. Each dimension's raw value is normalised to a 0–1 scale: (value − min) / (max − min).
  3. The normalised values are summed over the rubric's FULL set of dimensions (each weighted equally unless stated otherwise), divided by the rubric's total weight, and multiplied by 100. A dimension that has not been evaluated counts as zero — absence of evidence is not evidence of merit.

Because the arithmetic is identical everywhere, a score of 80 means the same thing in any catalogue: the practice meets, on average, 80% of its rubric's evaluation criteria. Catalogues differ only in which dimensions they assess, because their source data differs.

Reading a practice's score

A single number hides a lot. Most practices score high (often 90+), so the headline figure alone barely separates them. Two things on each practice page put the score in context.

  • The score profile shows the per-dimension breakdown behind the score, so you can see where a practice is strong or weak rather than just its average.
  • The catalogue position (e.g. “Top 10%”) shows where the practice sits among scored peers in the same catalogue. We count how many peers score at or below it (an “at or below” tie convention). It is shown only when a catalogue has at least five scored practices, so a position is meaningful.

Both are ways of reading the score — they never change the stored value. The score itself is always computed exactly as described above.

Dimensions assessed, by catalogue

AI in Education

Evaluated by C-NAPSE

DimensionScaleWeight
Evidence of learning impact0–31
Equity & inclusion0–31
Transparency, ethics & data protection0–31
Teacher capability & pedagogy0–31
Scalability & sustainability0–31

AI in the Public Sector

Evaluated by C-NAPSE

DimensionScaleWeight
Evidence of impact / measured public value0–31
Transparency, fairness & accountability0–31
Transferability / demonstrated replication0–31
Scalability beyond pilot0–31
Governance, capability & sustainability0–31

City Innovation Library

Evaluated by C-NAPSE

DimensionScaleWeight
Evidence of impact / measured results0–31
Transferability / demonstrated replication0–31
Scalability0–31
Sustainability / continuity0–31
Innovation0–31
Inclusiveness / equity0–31
Multi-stakeholder collaboration0–31

Digital Inclusion

Evaluated by C-NAPSE

DimensionScaleWeight
Innovation level0–51
Sustainability (ongoing)0–11
Evaluation evidence0–11
Demonstrated reach0–11
Documented learning0–11

Digital Inclusion

Evaluated by MEDICI consortium

DimensionScaleWeight
Innovation level0–51
Sustainability (ongoing)0–11
Evaluation evidence0–11
Demonstrated reach0–11
Documented learning0–11

Gender Equality

Evaluated by ProPEGE consortium (Equal Leadership)

DimensionScaleWeight
Transferability / replicability0–31
Impact on gender equality0–11
Effectiveness0–11
Efficiency0–11
Evaluated outcomes0–11
Sustainability0–11
Achievement / evidence0–11
Gender-mainstreaming embedding0–11
Curator validation0–11

Gender Equality

Evaluated by C-NAPSE

DimensionScaleWeight
Transferability / replicability0–31
Impact on gender equality0–11
Effectiveness0–11
Efficiency0–11
Evaluated outcomes0–11
Sustainability0–11
Achievement / evidence0–11
Gender-mainstreaming embedding0–11
Curator validation0–11

PES & Nature-Based Solutions

Evaluated by C-NAPSE

DimensionScaleWeight
Carbon — sequestration / storage, evidenced0–31
Biodiversity — conservation / improvement, evidenced0–31
Water & soil — retention, infiltration, erosion control0–31
Fire resilience — risk reduction, evidenced0–31
Governance, certification & transferability0–31

One canonical spine, many sources

The rubric method above is the canonical spine of every score on this platform. Several catalogues began life by importing practice collections built by earlier projects and consortia (for example ProPEGE for gender equality, MEDICI for digital inclusion). Their original evaluations are respected as attributed priors: where a source consortium assessed a practice, that assessment seeds the corresponding rubric dimensions and the source is credited on the practice page — but the dimensions, the arithmetic and the published score are always Evidoria's. Where something came from never decides how it is organised or ranked: provenance is metadata, recorded per record — never structure.

Validation levels

Every practice page shows how the record was produced and checked. The ladder, from most to least validated:

How our evidence ladder maps to other standards

Evidoria classifies each practice's evidence by study design. The table below maps that ladder onto three frameworks widely used by what-works units — the Maryland Scientific Methods Scale, Nesta's Standards of Evidence and EMMIE. Correspondences are indicative (a specific study can sit higher or lower), and EMMIE is multi-dimensional: only its Effect dimension maps onto a design ladder — its Mechanism, Moderators, Implementation and Economics readings correspond to the implementation dossier on each practice page.

Evidoria evidence strengthMaryland SMSNestaEMMIE
Randomised controlled trialLevel 5 — randomised assignment to treatment and controlLevel 3–4 — causal impact demonstrated; 4 where independently replicatedStrong Effect evidence — direct estimate of impact with a credible counterfactual
Quasi-experimentalLevel 3–4 — comparison group without full randomisationLevel 3 — causal comparison against a control or matched groupModerate-to-strong Effect evidence — counterfactual present, selection risks remain
Observational / pre–postLevel 2 — before/after measures without a comparison groupLevel 2 — data showing change among recipientsWeak Effect evidence — change observed, attribution not established
Descriptive / self-reportedLevel 1 — correlation or descriptive account onlyLevel 1–2 — a logical account of impact, possibly with descriptive dataNo Effect estimate — EMMIE's Mechanism/Implementation reading may still apply

Where the data comes from

Every practice lists its data sources — the catalogue or website each piece of information was retrieved from — together with the date it was accessed. You'll find these on each practice's page under “Data sources”.