Performance and confidence are not the same finding

An evaluation that reports one number is hiding the more useful one. How strongly a finding can be proven is a separate question from how well a programme performed.

Most evaluation reports hand over a single score, and a single score forces two different questions into one answer. How well did the programme perform? And how much weight can that answer actually bear? A programme can look strong on thin evidence, and a programme can look mediocre on a record so well documented that the finding will survive any scrutiny a donor brings to it.

The SSII Framework, developed by Samwel Ochieng Onono and applied in CASA NEXUS evaluations, reports the two separately and never one without the other: a Strategic Integrity Score for how the system performed, and an Evidence Confidence Index for how strongly that finding can be proven. The two are scored on the same weights and are never blended.

The evidence band then sets a claim licence, which fixes the strongest verb every product of the evaluation is permitted to use. Where the record supports it, the report says caused. Where four rival explanations survive testing, it says the evidence supports. That is not hedging; it is the difference between a claim that holds in front of a board and one that collapses the first time it is checked.

Protection is handled the same way, as its own dimension rather than a footnote to delivery. A safeguarding failure caps the reportable band regardless of the composite, the cap cannot be waived by a client, and it is lifted only by re-collected evidence. In practice, averaging is exactly how protection failures stay buried.