Program value waterfall theo at risk → saved → costs → net EBITDA

{{ vtWaterfall }}

projected (hatched) backtested (solid) net protected EBITDA Finance-auditable by construction

Cumulative lift · treated vs stratified holdout net revenue / guest · rolling 12 wk

{{ vtLift }}

Response latency: sustained change is judged quarterly, not next-day. Vertical markers = campaign waves.

Top considerations this period auto-generated · each bullet links to its evidence

{{ n.n }} {{ n.text }} Drill to evidence →

Open pilot decision pack

All ADT bands400–750750–2,5002,500+ All statesNew this weekIn progressActionedOutcome loggedExpired sorted by expected incremental value (uplift-ranked)

Sleeping dogs (8–12% of the cohort) never enter this queue — they are visible only in the Guardrails Console with a suppression reason. No bulk actioning: per-guest craft workflow is the compliance posture.

Assign to host Log outcome View full 360 Suppress… Every click is recorded in the append-only audit trail · “Assign” is SIMULATED until pilot

Theo trajectory · 24 months · changepoints + event context {{ eventsBtnLabel }}

{{ theoChart }}

Dashed verticals = structural changepoints (Dec tax change, net-ADT restructure). Shaded = AFL / school holidays / Crown Live — a footy spike is never misread as organic growth.

Offer history ledger · P&L after tax — select a row for its waterfall

Elasticity & context · modeled reinvestment response

RFM × ADT explorer All ADT bandsADT 400–750ADT 750–2,500ADT 2,500+ All distance rings< 30 km30–100 km100+ km

0–7d8–14d15–30d31–60d60d+

cell color = total theo · label = guest count · recency (cols) × frequency (rows). Click a cell to inspect the cohort.

Cumulative incremental net revenue · treated vs stratified holdout {{ eventsBtnLabel }}

{{ measLift }}

Response windows drawn as lag fans, not day-one deltas · headline = rolling 12 wk · regime legend: post-Dec-tax training window

Wasted spend saved · before vs after uplift targeting

{{ w.seg }} {{ w.saved }} saved · revenue held {{ w.held }}

before

{{ w.before }}

after

{{ w.after }}

Reference engagements saved ~60% of spend for the same outcome — “stop paying uninfluencibles.”

Offer P&L waterfall {{ c.label }}

{{ campWf }}

{{ campNote }} · mirrors the P&L view the CRM team already trusts in Power BI, deliberately, for adoption.

Stratification: ADT band × lifecycle stage × locality ring. A holdout guest who receives an offer through a legacy channel is flagged and excluded from the claim.

Rendered recommendations passing the filter chain

100% 2,847 evaluations · wk of 19 Jul

Self-exclusion, pre-commitment, consent and frequency caps are applied before a recommendation is rendered — compliance is architectural, not asserted.

{{ s.n }}

{{ s.reason }}

Export for audit

Mindway Game Scanner integration

{{ b.band }}

{{ b.n }}

{{ mindwayPolicy }} Displayed as bands — raw scores are role-restricted.

Language governance — centrally managed copy library

APPROVED FRAMING

{{ t }}

BANNED — never rendered

{{ t }}

Translation slots: {{ langTrans }}. Every word rendered assumes a future regulator will read it.

Draft offer

{{ pinLabel }} Export review summary Promote to Offer Bank Reset panel

Nothing in the simulator schedules or sends anything. Promoting attaches the panel summary as design evidence; scheduling runs through the normal flow with a holdout auto-reserved.

{{ s.label }}

{{ s.value }}

{{ s.sub }}

{{ grp.label }} {{ grp.count }} of 16 {{ grp.hint }}

{{ p.name }} {{ p.bio }}

“{{ p.rationale }}” {{ p.flipLabel }} {{ p.chev }}

Attribute sheet · synthetic composite

{{ dr.k }}{{ dr.v }}

Not a real guest — generated from cohort statistics in the feature mart (calibrated to 12 Jul). Fit radar in the panel below.

Drift, changepoints & retraining timeline

{{ e.w }} {{ e.label }}

Stated, on-screen behavior — never silent degradation: after the net-ADT restructure, the decline model temporarily widens its uncertainty and the queue shows fewer, higher-confidence flags for ~2 weeks. The changelog posts to the Value Tracker narrative card in plain language.