US Brand Signal Policy Simulator

US Brand Signal Policy Simulator compares several buyer-defined threshold policies over the same opaque signal scores. It returns the accepted, review, and rejected IDs for every policy, pairwise difference counts, and a per-signal sensitivity summary. It does not choose a winner, recommend a policy, verify a score, or make a truth claim.

What you submit and receive

You submit 1 to 100 unique rows such as {signalId, normalizedScore} and 1 to 20 unique policies with reviewThreshold < acceptThreshold. Scores and thresholds are integers from 1 to 100. The Actor sorts IDs and policy IDs, so input order does not affect the report.

You receive one Dataset report containing:

• policyResults: sorted acceptedIds, reviewIds, rejectedIds, counts, and a classification digest for every policy;
• comparisons: stable pairwise counts of changed and unchanged decisions, without ranking either policy;
• sensitivitySummary: how many policies put each signal in each outcome and its decision spread;
• SHA-256 digests for input, rows, policies, results, comparisons, sensitivity, and the complete report;
• a truthful OUTPUT receipt after one confirmed Dataset write.

Run it

1. Open Try for free in the Actor Console.
2. Paste the exact JSON from the Input section or examples/input.json.
3. Start the run and read the single report row from the Dataset.
4. Read the terminal delivery receipt from OUTPUT.

public-task.json is the exact public-task input and is identical to the runnable prefill.

Pricing

Pay per event: $0.005 per run start plus $0.010 per delivered result on the FREE tier. The result event is emitted only after one confirmed Dataset write. The buyer cap must cover both events.

Tier	Discount	Start	Delivered result
FREE	0%	$0.00500	$0.01000
BRONZE	5%	$0.00475	$0.00950
SILVER	10%	$0.00450	$0.00900
GOLD	15%	$0.00425	$0.00850
PLATINUM	18%	$0.00410	$0.00820
DIAMOND	20%	$0.00400	$0.00800

Invalid input, insufficient cap, uncertain delivery, and an uncharged or multiply charged result are not reported as successful results. The Actor never retries a paid Dataset push.

Freshness and determinism

There is no external freshness: this Actor reads only the JSON supplied in the run and does not read the clock. Rerun when the buyer's upstream scores or policies change. The same input always produces the same sorted rows and digests.

Input

{
  "schemaVersion": "1.0",
  "rows": [
    {"signalId": "signal-001", "normalizedScore": 96},
    {"signalId": "signal-002", "normalizedScore": 82},
    {"signalId": "signal-003", "normalizedScore": 61},
    {"signalId": "signal-004", "normalizedScore": 39}
  ],
  "policies": [
    {"policyId": "balanced", "reviewThreshold": 50, "acceptThreshold": 80, "thresholdOrder": "review_lt_accept"},
    {"policyId": "strict", "reviewThreshold": 70, "acceptThreshold": 90, "thresholdOrder": "review_lt_accept"},
    {"policyId": "broad", "reviewThreshold": 30, "acceptThreshold": 70, "thresholdOrder": "review_lt_accept"}
  ]
}

Output

The Dataset has exactly one report item when delivery succeeds. policyResults are sorted by policyId; each outcome ID list is sorted by signalId. comparisons use the stable policyA < policyB order. OUTPUT contains the terminal delivery receipt and links represented by the Store output schema.

Related public Actors

These Actors can feed or use the same buyer-owned signal workflow. This Actor does not call them.

Actor	Use it for
US Brand Offer Evidence Normalizer	Normalize evidence-derived inputs before simulating threshold policies.
US Brand Signal Metrics Aggregator	Summarize score rows before policy simulation.
US Brand Action Queue	Convert a buyer-selected outcome into next-action rows.
US Brand Signal Brief Composer	Turn a bounded comparison report into an agent-readable brief.
US Brand Evidence Snapshot Diff	Compare two buyer-supplied snapshots before updating scores.

Limitations

• IDs and scores are opaque buyer input. Accepted does not mean verified, true, safe, valuable, or likely to convert.
• There are no network requests, proxies, BYOK keys, LLM calls, clock reads, state, child Actors, or Dataset input.
• The maximum input is 100 rows and 20 policies. Duplicate IDs, extra fields, non-integer values, and invalid threshold ordering fail closed.
• A complete report must fit below 64 KiB. If a valid maximum-size input would exceed this bound, the Actor returns output_too_large instead of truncating or charging for a partial report.
• JSON Schema validates structure; runtime validation additionally enforces sorting, digests, counts, threshold ordering, and exact report equality.

FAQ

Does the Actor choose the best policy? No. It only shows outcomes and differences. The buyer or downstream workflow decides.

What does sensitivity mean here? It is the number of submitted policies that place one signal in accepted, review, or rejected, plus the number of distinct outcomes.

Are scores probabilities or verified evidence? No. They are buyer-supplied normalized integers used for deterministic simulation.

What happens at a boundary? Accept is inclusive at acceptThreshold; review is inclusive at reviewThreshold and exclusive at acceptThreshold; lower scores are rejected.

What happens after uncertain delivery? The Actor makes no second Dataset push and records a delivery error instead of claiming success.

Built by zinin.