n.V3-3.09 | MEASURING WHETHER IT HELD

01| Everything in this book part rests on one claim: IkoCiti's impact does not evaporate. If that is false, the differentiation collapses and we are an unusually well-designed version of what already exists. So anti-evaporation measurement is not a reporting function. It is the test of the thesis.

02| Most organisations do not run it, and the reason is not laziness. Measuring evaporation means following people for years, which is expensive; being honest about outcomes, which is disappointing; and accepting that completion is not success, which is professionally inconvenient. Not knowing is comfortable. We have made not knowing structurally unavailable to ourselves, and we should be clear that this will produce bad news in public.

03| Impact evaporates through five distinct mechanisms, and they need different responses.

04| Capability decay. Skills fade without use — fast at first, then slower. This is not speculation; it is measured. Health workers trained in neonatal resuscitation in South Sudan scored 97.8% on knowledge immediately after training and 69.4% twelve months later; their bag-and-mask skill fell from 94.4% to 77.0%. A comparable Ethiopian study found median knowledge falling from 98% to 74% in a year. Decay is normal. The question is whether our approach decays slower than the alternatives and whether refreshers reverse it.

05| Application opportunity loss. A Maveriq retains the capability but loses the situation in which to use it. Jobs disappear, family circumstances change, health intervenes. Partly outside our control — but capability breadth and ongoing connection to opportunity create multiple pathways where a single-skill training creates one.

06| Support withdrawal. Gains that depended on a coach, a mentor or a cohort disappear when the support does. Our peer model is supposed to make support structural rather than programmatic. That is a design intention, and it needs verifying rather than asserting.

07| Crisis disruption. Poverty involves crisis. A health emergency, an eviction, a legal problem — each can erase months of progress. We cannot prevent crises. We can measure whether Maveriqs recover faster than they used to, which is the only meaningful definition of resilience.

08| Context defeat. The individual develops and the environment does not move. Skills with no local labour market, financial management with no income to manage, confidence against a wall of discrimination. This is the limit of any individual-focused intervention, and our answer — that enough Maveriqs shift the environment itself — takes years. Individuals will hit context defeat before the community effect arrives.

09| We measure across four windows: 0–3 months, 3–12 months, 1–2 years, 2–5 years. Early loss means the transition was unsupported or the capability was never deeply acquired. Loss at one to two years is where most programmes stop looking. Persistence at five years is the only result that settles the argument.

10| Five things get tracked: capability retention, outcome persistence, support connection, crisis response, and progression versus stagnation. The last one carries the most weight — the share of Maveriqs who go on to train others is the ultimate indicator, because it is the loop closing.

11| Thresholds have to be set before the data arrives, or every result gets reframed as acceptable. Under 10% loss at twelve months is exceptional. Ten to 30% is normal decay we can work on. Thirty to 50% is a problem demanding intervention. Above 50% is failure, and we would have to say so.

12| Against the retention literature, our working expectation of 60–80% capability retention at twelve months is defensible but not comfortable. It sits roughly where trained health workers land — and those are people with formal supervision and on-the-job reinforcement. We are betting that continuous project work is a stronger reinforcement than periodic refresher training. That bet is testable and we should treat a miss as information, not as an accusation.

13| Numbers without comparison mean nothing. Seventy percent retention is good or bad only against something: published benchmarks from similar programmes, our own cohorts against each other, and — the strongest available — dose-response, whether Maveriqs with more platform engagement retain more. Perfect attribution needs randomisation we will not have. Self-selection affects every comparison we can run. Overclaiming will cost us more credibility than admitting the limit.

14| Measurement earns its cost only when it changes something. Loss concentrated in the first three months means the transition needs support. Loss following crises means resilience training and a crisis protocol. Loss correlated with low trainer contact means the contact requirement was too soft. Specific patterns, specific responses.

15| The organisational problem is patience. A twelve-month retention rate takes twelve months. By the time we know what the current training approach produced at twenty-four months, the approach will have changed — so we have to track which version of ourselves produced which result. And the commitment has to outlast the people who currently understand why it matters.

16| Measurement that produces only good news is not measurement. The credibility of everything in this volume depends on our willingness to publish the number that says we were wrong.