RG Interventions Playbook

The ladder

Rung	Interventions	Matched to
1 · Ambient	Visible session/spend data, reality checks, always-on tool access	Everyone — the floor the product ships with
2 · Prompted	Pattern-based check-in messages, suggested limits, break prompts	Early markers: baseline drift, mild escalation
3 · Conversation	Trained-handler contact with a documented structure	Sustained markers, chasing patterns, self-signals
4 · Imposed	Mandatory limits, cool-offs, account review, promo removal (already automatic at rung 2+)	Behaviour continuing through rungs 2–3, or severe markers directly
5 · Exclusion support	Supported path to self-exclusion, register signposting, external help referral	The player's decision, or the regime's mandated cases

Two structural notes: promotional suppression is not a rung — it happens automatically the moment a case opens, per the operations page; and skipping rungs is a written rule for severe marker classes, not a handler's improvisation.

Tone: the craft rung 2 and 3 live on

The message rules that survive both player research and regulator review: lead with the player's own data, not a policy; no moralising and no euphemism; one specific offered action per message (set a limit, take a break, talk to someone); and a one-tap path to a human. Templates are versioned and reviewed like the compliance surface they are — and generated variants, where used, operate inside the locked-template discipline from AI CRM.

Timing beats intensity

The consistent operational finding across this Academy's queue disciplines applies here with the highest stakes: early and light outperforms late and heavy. A rung-2 prompt inside the session where the pattern emerged does more than a rung-4 imposition weeks later — which is why marker-to-intervention time is the function's core SLA, and why the detection page insists on baseline-relative markers that fire early. The playbook's job is to make the early step so cheap and so normal that it actually happens.

The case trail through the ladder

Every step logged against the case: what was sent or said, the player's response, the tool taken or declined.
Every escalation shows its criteria: the written rule that moved the case up a rung, not a handler's mood.
Every closure has a state: resolved with behaviour change, resolved with tools in place, escalated, or no-action-with-rationale — each with a review date.
The trail reads as a story: a regulator opening one case file should see observation → proportionate response → escalation logic → outcome. That readable story, multiplied by every case, is the audit position.

Continue reading: Reporting and audit — the evidence the ladder produces. Detection markers — what opens the case.

Interventions — common questions

What makes an intervention proportionate?

Matching the response to the signal strength, with escalation when the lighter step does not change the behaviour. The ladder runs from ambient information (visible session data, reality checks), through active friction (prompted breaks, suggested limits), to protective conversation (a trained human, not a script), to imposed controls (mandatory limits, cool-offs) and finally supported exclusion. Proportionality is what makes the early rungs credible — a function that jumps to heavy steps teaches players to hide.

Does tone actually matter, or is that soft talk?

It is the difference between an intervention and an accusation, measurably: check-ins framed around the player's own patterns ('your sessions have been running longer than usual — here is what that looks like') get engagement; compliance-flavoured messages get dismissal and, worse, concealment. The writing rules are concrete — the player's data, no moralising, a specific offered action, an easy human path — and templates are reviewed with the same seriousness as compliance wording, because they are compliance wording.

Who should deliver the human conversation?

Trained RG handlers — not the VIP manager whose bonus depends on the player's turnover, and not general support reading a script. The conversation has a documented structure: what was observed, what tools exist, what the player chooses, what follows if nothing changes. It is recorded in the case trail like every step, and the handler has the authority to impose the next rung if the conversation makes the risk clearer.

When do voluntary tools become imposed ones?

When the behaviour continues through the voluntary rungs, or when a marker class is severe enough to skip them — both written in the playbook, not judged per case. Imposed steps (mandatory limits, forced cool-offs, account review) carry their own rules: clearly communicated, never punitive in tone, with the rationale logged and the appeal path stated. The escalation criteria being written is what protects both the player and the handler from inconsistency.

How is intervention effectiveness measured without gaming it?

Behaviour change on matched windows (the markers that fired, quieting), tool uptake and retention of tools over time, repeat-escalation rates, and — the one that keeps the system honest — no revenue metric in the objective. Reported beside, per the operations page, never optimised. An intervention programme tuned for revenue retention will drift toward theatrical light-touch steps; the measurement design is what prevents that drift.

The interventions ladder.

The ladder

Tone: the craft rung 2 and 3 live on

Timing beats intensity

The case trail through the ladder

Interventions — common questions

See it on a live stack.