Pilot metrics and impact reporting
Three-month pilot · 1 April 2026 – 30 June 2026 · single arm, no control group. Everything below is aggregate and de-identified.
Read these three conditions before using any figure on this page
- De-identified. Behaviour logs are stored without identifiers. This page has no search by person, no profile view, and no path back to an individual account.
- Minimum cohort size of 10. Any group smaller than 10 people is not displayed, even when the total is large enough — small groups can be re-identified from the inside.
- Self-reported figures are a proxy indicator, not an objective measurement. They must not be presented as evidence that the product works.
METRIC-01 · Assessment score, before and after
n = 41 matched pairsA standalone knowledge assessment sat at the end of onboarding and again two weeks after each participant's baseline. Scored out of 20.
A difference of +3.8 points (+19 percentage points). This is change within a single group. With no control group it cannot be stated as “the app increases knowledge” — practice effects from sitting the same instrument twice, and self-selection into the pilot, have not been ruled out.
Breakdown by age band
| Age band | n | Baseline | Endline | Change | Status |
|---|---|---|---|---|---|
| 18–24 | 18 | 10.8 | 15.0 | +4.2 | displayed |
| 25–34 | 14 | 12.1 | 15.6 | +3.5 | displayed |
| 35–44 | 6 | — | — | — | not shown · n < 10 |
| 45 and over | 3 | — | — | — | not shown · n < 10 |
The 9 people in the two suppressed bands are still counted in the total of 41 above. Only the band-level detail is withheld.
METRIC-02 · Return rate at day 1, 7 and 30
base sample 64Retention is not a success measure for this product and must not be treated as one. For a gambling-harm tool, high retention can be a bad sign. Read it only alongside METRIC-03 and METRIC-04. A report that leads with retention is reporting against the wrong goal.
METRIC-03 · Self-report survey (opt-in)
PROXY INDICATORPeriodic questions about real gambling outside the app. Entirely voluntary; declining costs nothing and changes nothing about the experience. Responses are stored de-identified.
The only permitted way to present this figure is “38 people reported the following”, never “the app reduced real gambling for 39% of users”. These are self-reported answers, subject to social desirability bias, collected from a group that chose to take part.
The “more than before” group is n = 5, so it must not be broken down any further on any dimension. As it stands METRIC-03 is not sufficient to say whether a gateway effect exists or not; answering that question would take an independent study design.
METRIC-04 · Continued after a risk warning
primary behavioural measureCounts how often the risk-warning modal was shown, and how often people went ahead anyway after reading it. Measured in weeks of app use, not calendar weeks. 1,842 warnings shown in total.
A fall of 19 percentage points over six weeks is the most interesting signal of the period. However, only about a third of the original sample is still present by week 6, so the decline may simply reflect that the people most easily pulled in left earlier. Do not read this chart as causal.
Weeks 7 to 12 fall below the minimum cohort size in each individual week and are therefore not plotted.
METRIC-05 · Learning to recover, or resetting
244 decision pointsWhat people chose at the branch points: losing an asset, reaching a critical state, and the end of a run.
The third recovery path is not built yet, so this period has no three-way comparison. The group that left without choosing is the one worth watching most and the one we know least about — they are no longer in the sample to be asked.
METRIC-06 · Support links opened
214 opens · 37 people| Where the link sits | Opens | Share |
|---|---|---|
| Critical state screen | 68 | 32% |
| End of run screen | 52 | 24% |
| Seven-day cooldown | 44 | 21% |
| Footer and support centre | 31 | 14% |
| AI companion | 19 | 9% |
This is a count of opens only. We do not record who tapped, we do not store companion conversations, and nothing is passed to the helpline. An open does not mean the person called or was helped — this measures reach, not outcome.
METRIC-07 · Three-month pilot report
A snapshot of the period's figures is generated when the period closes, and cannot be edited afterwards, so later periods have something fixed to be compared against.
Quantitative success criteria for the pilot have not been set. This first period exists to establish a baseline, not to clear a threshold.
Limitations of this pilot — these travel with every external presentation of the figures
- Single arm, no control group. There is no comparison group who did not use the app, so no before-and-after change here can be attributed to the product.
- Self-selected sample. People who enrolled in the pilot may already have been interested in gambling harm. The results do not generalise to the wider population.
- A sample of 64 over three months is too small to detect a real effect, particularly within subgroups.
- Attrition. Only 41 of 64 have a complete baseline–endline pair, and people who left early are barely measured at all. Later metrics always lean towards those who stayed.
- METRIC-03 is self-reported and cannot be checked against actual betting behaviour outside the app.
- Commitment to publish. Negative or neutral results are published in full. If the data suggests people bet more after using the app, the pre-registered response is to pause or change the product and publish that finding.
What must not be done with this data: read it back into any individual person's experience; use it to personalise odds, rewards or notifications; or let it imply the product helps anyone win real bets.