Pilot report for the funding body
Everything tibbi sends to the ACT Gambling and Racing Commission is assembled here: the progress reports during the grant term, the three-month pilot report, the public summary, and the final report and acquittal. Each one is built from a closed period snapshot and from nothing else.
This report is assembled, not written
- No figure on this page can be typed in. Every number is read from a closed snapshot, so any figure the Commission receives can be checked back against the period it came from. Open the snapshots →
- The limitations block cannot be switched off, shortened, moved to an appendix or left out of the public summary. It travels inside the same document as the numbers.
- A released report cannot be edited. A figure found to be wrong is corrected by issuing a correction beside the original, with both left on the record.
What the Commission receives, and when
| Document | Built from | Due | State | |
|---|---|---|---|---|
| Progress report 1 internal trial, before the pilot opened |
PS-2026-Q1 | 15 Apr 2026 | Released 10 Apr 2026 | Read as released |
| Three-month pilot report the report this whole subsystem exists to produce |
PS-2026-Q2 (not yet generated) | within 30 days of the pilot closing | Draft · assembles when Q2 closes on 31 Jul | Open the draft |
| Public pilot report PDF and a plain summary page anyone can read |
the released three-month report | 30 Aug 2026 | Waiting on the report above | See the summary |
| Final report and acquittal completion of the project and acquittal of the funding |
every snapshot in the grant term | within 30 days of the end of the term | Not yet due | No action available |
Reports are cumulative and none replaces another. The final report includes the progress reports as they were issued, including any figure a later period turned out to contradict.
Released · Progress report 1
What this report did not have
Both slices the Commission asks for in every reporting period were only half answerable in the trial, and the report said so rather than leaving the heading out:
- Residence: 11 of the 12 gave an ACT postcode. Published as a count, self-declared and unverified.
- Where people started: not collected. The screen that asks whether someone has never bet, is currently betting, or is cutting down went live with the pilot on 1 April 2026. The trial has no answer and none was estimated.
A heading with “not collected” under it is a report. A heading quietly removed is not.
The document, as it was sent
read-onlyInternal trial, 1 January – 31 March 2026. Twelve people, nine of them working on the product. The trial tested whether the measurement plumbing works, not whether the product helps anyone.
- Knowledge before and after: no result. Twelve baseline sittings, no endline sittings, no matched pairs. The endline instrument was not in field during the period. No knowledge gain is claimed from this period because there is not one to claim.
- Risk warnings: 152 of 214 shown warnings were read and the action taken anyway. Reported as a count. The trial is far too small to publish this as a rate.
- Support links: 21 opens by 6 people. A count of opens measures reach. It does not show that anybody was helped and it never will.
- Returning: 10, 8 and 5 of 12 at day 1, 7 and 30. Returning is not a success measure for a gambling-harm product and is not offered as one.
Limitations, reproduced from the released document. Twelve people, most of whom built the product, cannot tell us whether it helps a stranger. There is no control group, no comparison, and no measurement of real gambling behaviour in this period. Nothing in this report should be read as evidence that tibbi works. Its only claim is that the measurement machinery records what it says it records.
In draft · three-month pilot report, Q2 2026
Draft. Not for release and not for quoting. The Q2 2026 period is still open and closes on 31 July 2026. Every figure below is read from the live dashboard, moves at 06:00 each night, and is replaced by the locked snapshot figure the moment the period closes. The release control at the bottom of this page stays off until then.
1 · Who the pilot reached
Reported in every period, because it is the evidence that the funding reached the people it was granted for.
Residence · 64 people in the measurement sample
ACT residents 51 of 64 · 80%Residence is a postcode the person typed at sign-up. It is not verified against anything, and the report says so wherever the figure appears. Nine people in surrounding New South Wales is what a border region looks like; they are counted and shown, not quietly dropped to make the ACT share read higher.
64 of the 96 registered accounts are in the measurement sample. The other 32 either declined the measurement screen at the end of onboarding or registered outside the period. They use the product in full, and they appear in no figure in this report.
2 · Where people started
Self-declared once, at the end of onboarding. Without this split the whole sample is averaged into a single number and the group the Commission most needs to hear about — people who had never bet with real money before using a product about betting — disappears into it.
Risk stratification at the start of the pilot
n = 64| Group | People | Share | Reported as |
|---|---|---|---|
| Had never bet with real money | 21 | 33% | Figures published |
| Currently betting with real money | 34 | 53% | Figures published |
| Cutting down or stopped | 9 | 14% | Below the minimum cohort size · described in words, no figures |
The nine people cutting down are counted in every total on this page and broken out in none of them. Nine people who have told us they are trying to stop is exactly the group a published figure could identify from the inside, and the report says why the row is blank instead of leaving the row out.
Two-directional gateway questions · T0 and three months
Not enough data yetThe same set of questions is asked at sign-up and again three months later, and it is written to catch movement in both directions — more real betting as well as less. The comparison needs both sittings from the same person.
| Stage | Collected | State |
|---|---|---|
| T0 · at the end of onboarding | 64 of 64 | Complete |
| Three months · due 1 Jul – 30 Sep 2026 | 11 of 64 | Still being collected |
| Usable pairs | 11 | Below the minimum cohort size · no figures published |
Eleven pairs is what exists, and eleven is what is printed. There is no chart here, because a chart of eleven people drawn on the same axes as a chart of sixty-four invites a reader to treat them as the same kind of evidence. The gateway question is answered in the report that covers the period the responses actually fall in, and not before.
3 · Headline figures, each with what kind of number it is
The working for each measure sits on the pilot metrics page. This is the report altitude: the figure, what it is, what it rests on, and the limit that is printed with it every time it appears.
| Measure | Figure | Kind of number | Sample | Printed with the figure, always |
|---|---|---|---|---|
| Knowledge, before and after METRIC-01 |
11.4 → 15.2 of 20 +3.8 |
Measured | n = 41 of 64 enrolled · 64% people with a baseline and an endline on the same version |
Single arm, no control group. This change cannot be attributed to the product. Sitting the same instrument twice produces some of it, and the 23 people without a pair are not represented in it at all. |
| Returning at day 1, 7, 30 METRIC-02 |
72% · 45% · 31% | Measured | 64 | Not a success measure. For a gambling-harm product, high retention can be a bad sign, and this project does not report against it as a goal. |
| Self-reported real betting METRIC-03 |
15 less · 18 no change · 5 more | Self-reported proxy | 38 of 64 opted in | “38 people reported the following”, never “the app reduced gambling”. The group reporting more is 5 people and is not broken down any further. |
| Continued after a risk warning METRIC-04 |
68% of warnings | Measured | 1,842 warnings | The fall across six weeks is reported with the caution that only about a third of the sample is still present by week six. |
| What people chose after a loss METRIC-05 |
132 of 244 chose learning | Measured | 244 decision points | The third recovery path is not built yet, so this period offers no three-way comparison. The 41 who left without choosing are the group we know least about. |
| Support links opened METRIC-06 |
214 opens · 37 people | Measured reach | 64 | A count of opens. Nothing is passed to the helpline, so the report cannot and does not say anyone was helped. |
Full working, breakdowns and charts for each measure →
Knowledge change by where people started
41 matched pairs| Group | Pairs | Baseline | Endline | Change | State |
|---|---|---|---|---|---|
| Had never bet with real money | 14 | 12.6 | 16.1 | +3.5 | published |
| Currently betting with real money | 24 | 10.6 | 14.6 | +4.0 | published |
| Cutting down or stopped | 3 | — | — | — | not shown · n < 10 |
All three groups are inside the 41 pairs. Both published rows still carry the same limit as the headline: no control group, so neither number can be attributed to the product. The two rows are shown separately because they are two different questions — whether a person who already bets understands the odds better, and whether a person who has never bet has been taught something they will use.
4 · Where there is not enough data yet
Every one of these is printed in the report with the counts that exist. None is estimated, extrapolated, or drawn as an empty chart with a hopeful axis.
- Three-month gateway comparison — 11 pairs of 64. Responses are due to 30 September 2026. Reported in the period they fall in.
- People cutting down — 9 of 64. Below the minimum cohort size. Described in words, with the count, and no figures.
- Age 35–44 (6 people) and 45 and over (3 people). Suppressed at every level including in the export, and still counted in every total.
- Any comparison with Q1 2026. That period ran a twelve-person internal trial on a different version of the assessment. The two cannot be pooled and no trend line is drawn from them. See the comparison and why it refuses →
- Whether a gateway effect exists. Answering that takes an independent study design with a control group. This pilot cannot answer it, and the report says so in those words.
5 · Limitations of this pilot
This block travels with the numbers into every document, including the public summary
- 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 may already have been interested in gambling harm. The results do not generalise to the wider population.
- 64 people over three months is too small to detect a real effect, particularly within subgroups.
- Attrition. 41 of 64 have a complete baseline–endline pair, 9 passed the endline window and can no longer be paired, and people who left early are barely measured at all. Every later figure leans towards those who stayed.
- Self-reported figures are a proxy indicator and cannot be checked against real betting behaviour outside the app.
- Residence and starting point are self-declared and are not verified.
- No quantitative success criteria were set for this pilot. It exists to establish a baseline, not to clear a threshold that was chosen before any data existed.
The block is part of the report body, not an appendix. There is no control in this console that removes it, shortens it, or exports the figures without it.
6 · Commitment to publish
- Results are published whether they are favourable or not. This is policy, not an intention. A period that produces an unflattering figure is reported with that figure in it, because the period is closed and the figure cannot be changed.
- If the data shows people betting more after using the product, the committed response is to pause or change the product and to publish that finding, in the same period it appears in.
- The public version is a plain-language summary plus the full PDF, published within 30 days of the pilot closing, and it carries the limitations block in full.
- The measurement instruments are shared openly — the knowledge assessment and the self-report survey — so another service can reuse or challenge them.
Owner: impact analyst. Nothing in this list is conditional on the result being good.
The public summary page
One page, no jargon, for anyone who wants to know what the pilot found without reading the full report. It is generated from the released report and cannot be written separately.
Preview · what the public page will say when Q2 closes
draftWhat we did. 64 people in and around Canberra used tibbi for three months. tibbi is a game about betting in which nobody bets real money and nothing is paid out. It uses play money to show how odds and the house edge actually work.
What we found. People scored higher on a short knowledge test after two weeks than before — 11.4 out of 20, then 15.2 — for the 41 people who sat both. 38 people answered a voluntary question about their own real betting: 15 said less than before, 18 said no change, 5 said more. Two thirds of the risk warnings shown in the game were read and the action taken anyway, and that share fell over six weeks.
What we cannot say. We cannot say tibbi caused any of it. Everyone in the pilot used the app; there was no comparison group who did not. The people who joined were already interested enough to volunteer. 64 people is a small number, and 23 of them do not have a complete before-and-after pair.
Limitations, printed on the public page itself: single arm with no control group; a self-selected sample of 64; results that do not generalise; 41 of 64 with a complete before-and-after pair; self-reported answers that we cannot check; and no quantitative success target, because this pilot exists to set a baseline rather than to clear one.
The summary and the block above are one document. There is no version of the public page with the numbers and without the limitations, and no export produces one.
Release
Before this report can be released
| Condition | State |
|---|---|
| Q2 2026 closed | Not yet · 31 Jul 2026 |
| Snapshot PS-2026-Q2 generated | Follows the close |
| Both mandatory slices present ACT residents · starting point |
Present |
| Limitations block present in report and summary | Present · cannot be removed |
| No group below the minimum cohort size published | Checked · 3 groups suppressed |
| Every figure traceable to a snapshot | Live figures until the period closes |
Release is off until Q2 2026 closes on 31 July 2026 and its snapshot exists — a report released from figures that can still move is a report that can be contradicted by its own source. Releasing records who released it and when, and closes the document. There is no unrelease.
Who may do what here
- The impact analyst assembles and releases reports, and writes the analyst note that closes a period.
- Nobody edits a figure. The console has no field for it, so the question of who is permitted never arises.
- A system administrator can restore this console from backup and can do nothing about the content of a released report.
- The funding body receives documents; it has no account in this console and no way to ask for a figure to be reworked before release.
This page holds no personal data and produces no export that contains any. Data access requests from individuals are handled elsewhere, by a different role.