Other tools optimize the number your ad platform reports. MIZ OKI optimizes the number your bank account reports.
A nervous system for your business — media, commerce, site, inventory, margin, and policy sensed together, so every decision about the next dollar is diagnosed, proven, governed, and learned from.
Commercial teams are surrounded by data and still cornered by the same questions — asked in meetings, answered by intuition, owned by nobody in particular.
Every one of these questions resolves the same way: one governed decision pathway.
Growth teams do not need another chart. They need a defensible path from a performance change to the next authorized move.
Connect media, site, inventory, CRM, and margin evidence to rank what most plausibly explains the observed change.
Measure incremental contribution instead of treating platform attribution as causal truth.
Compare candidate moves against profit, risk, downstream effects, and no-action.
Apply policy, authority, approval, rollback, and audit controls before dispatch.
Every MIZ OKI Media decision moves through the platform's seven-stage operating loop — SRPVDAL — in this exact order, and every stage leaves a record. Nothing executes until it is validated against your constraints and authorized by the right person or the right delegated rule.
Detect meaningful commercial change — media, commerce, site, inventory, and economic thresholds, normalized into evidence with provenance.
Evaluate competing hypotheses for why performance moved, with the supporting and disconfirming evidence preserved.
Generate multiple possible interventions — including doing nothing — and compare their expected outcomes.
Apply organizational constraints: finance, margin, inventory, brand, compliance, and risk. Vetoes are recorded, never silent.
Route the recommendation to the correct authority — a named person, a delegated rule, or a bounded automation.
Execute only the authorized action — scoped, bounded, with its rollback path preserved. "No action" is a first-class outcome.
Compare predicted with realized outcomes and write the adjustment into memory, so the next decision starts from a smarter baseline.
Before a signal can travel the pathway it enters through one governed gateway, where every external source is normalized into a single canonical business event. Nothing goes from raw signal straight to action.
Mail & thread signals.
Documents & contracts.
Code, issues & CI.
Commitments & timing.
Warehoused facts.
Graph relationships.
Reasoning & embeddings.
Commerce, orders & revenue.
Search acquisition signals.
Analytics & conversion signals.
Facebook & Instagram signals.
Email delivery & engagement signals.
Programmatic bidstream signals.
Email & lifecycle marketing.
Programmatic display & video.
Retail media spend & reported sales.
Retail media spend & reported sales.
Deterministic policy, human-in-the-loop keys, and four deterministic decision gates screen every autonomous action before it touches the business.
Context-aware controllers suggest an operational action or capital movement.
Counterfactual simulations run the action against contracts and legal frameworks.
The decision gates apply deterministic policy and human-in-the-loop approval.
Secure, immutable execution with a tamper-evident audit trail.
Every proposed action is evaluated per decision, per domain, against four deterministic gates — identity, policy alignment, context sufficiency, and hard constraints (like a fiduciary floor). All gates must pass; any single breach — like ACT-991's — vetoes the action regardless of how strong the others are. ILLUSTRATIVE.
Adjust the parameters of a hypothetical proposed action and watch the decision gates evaluate it.
The 90-day pilot is the commercial standard: it starts observe-only and advances one rung at a time, on evidence about your own decisions. Most platforms sell the last rung on day one. Here, the sequence is the product.
A model that misses any threshold recommends. It does not act.
The thresholds are design targets, scored per account and per action class. Today the AUC gate is not yet scorable — the live tables hold zero admissible forward labels — so the intent cells observe. That gap is on this page on purpose.
THE FULL PILOT STRUCTURE — METRICS, CHECKPOINTS, EXIT CRITERIA →Every capability on this site carries one of four labels — LIVE, PARTIAL, IN BUILD, PROPOSED — and the classes are never mixed. The rows below are the platform's single status source, reproduced here rather than paraphrased.
| SRPVDAL loop · Decision Control Plane · DEL scoring · audit ledger | live |
| Connector gateway · canonical envelope · O-1 privacy lock | live |
| F4 micro-geo calibration (pilot armed 2026-08-24 · every reservation L2-approval-gated · bounded autonomy after 2 clean cycles, 0 complete) | live |
| Causal proof core · credit ledger · measurement rails | partial |
| ORACLE cells 33–36 · intent edges (observe-only / shadow) | partial |
| Intent Engine v2 · Net Yield lane · F1 F2 F3 F5 | in build |
| Ghost bids · clipped-ReLU authorization shape · ad control plane | proposed |
Every system you run remembers what happened. None of them remembers why you decided — what the evidence was, which causes were weighed, what policy allowed, who approved, what was done, and whether it worked. The MIZ OKI Decision Graph exists to remember exactly that: Signal → Entity → Cause → Objective → Policy → Action → Outcome, each stage a first-class object, with the realized outcome written back so the next decision starts from a smarter baseline. It is the Firestore-backed substrate every product surface reasons over — and the Growth Decision Graph is its view for growth teams.
Start with the product. The illustrative demos and the technical walkthrough show the same governed decision pattern on synthetic data — no signup or sales call required.
The product: Causal Growth Control for CMOs and performance marketers. The executive demo — five acts, one synthetic brand — is one click away inside it.
Six interactive desks apply the same decision pathway across domains on illustrative data. Open a desk, run a scenario, inspect how the decision trail assembles.
The guided tour of the operating loop, the decision gates, and the audit trail — for the people who will evaluate the technology.
Try each decision job hands-on in the Decision Studio — an illustrative walkthrough of the decision tree on synthetic data. No sign-in required.
Tell us the decision that costs you the most when it goes ungoverned. We scope a 90-day pilot around it, start observe-only, and put the first registered holdout in front of you by day 31.