US Brand Signal Confidence Gate

US Brand Signal Confidence Gate is a deterministic signal-scoring classifier for teams that already have normalized US brand signals. Give it 1-100 buyer-supplied scores and an explicit accept/review policy; it returns sorted accepted, review, and rejected signal IDs with closed reason codes, stable counts, and SHA-256 digests.

It does not verify the signal, fetch a source, identify a person or company, or claim that an accepted row is true.

What you get

• Exact acceptedIds, reviewIds, and rejectedIds lists sorted by opaque buyer signal ID.
• One decision row and one closed reason code for every input row.
• Explicit threshold ordering: reviewThreshold < acceptThreshold and thresholdOrder: "accept_gt_review".
• Reconciled counts plus input, row-list, outcome-list, and result digests.
• One bounded Dataset report and a truthful OUTPUT terminal receipt.
• Runs on Apify with JSON export, API access, schedules, monitoring, and pipeline integration after your own publication decision.

How to run it

1. Click Try for free in the Actor Console.
2. Provide thresholds and rows in the JSON input. Each row needs an opaque signalId and an integer normalizedSignal from 1 to 100.
3. Set acceptThreshold higher than reviewThreshold, keep thresholdOrder exactly accept_gt_review, and click Start.
4. Read the one report row from the Dataset and the delivery receipt from OUTPUT.

No source URL, proxy, API key, Dataset ID, identity field, or runtime timestamp is accepted.

Pricing

Pay per event: $0.005 per run start + $0.010 per delivered result on the 0% FREE tier. The result charge is emitted only for one confirmed Dataset write with the named result-found event. The six exact PPE tiers are:

Tier	Discount	Run 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

The runtime checks the buyer cap for the start plus one result before writing the Dataset. Invalid input, budget rejection, uncertain delivery, and uncharged delivery are not reported as successful results. One example batch of four rows costs $0.015 on FREE when delivered successfully.

Freshness and determinism

This Actor has no source freshness to measure because it never fetches data and never reads the runtime clock. Freshness is the buyer's responsibility: rerun when the upstream signal set changes. The same canonical input always produces the same report and digests, regardless of row order.

Input

{
  "schemaVersion": "1.0",
  "thresholds": {"acceptThreshold": 80, "reviewThreshold": 50, "thresholdOrder": "accept_gt_review"},
  "rows": [
    {"signalId": "signal-001", "normalizedSignal": 96},
    {"signalId": "signal-002", "normalizedSignal": 82},
    {"signalId": "signal-003", "normalizedSignal": 61},
    {"signalId": "signal-004", "normalizedSignal": 39}
  ]
}

The inline input above is the complete runnable fixture. A successful run returns the one report row described below and a terminal receipt in OUTPUT.

Output

The Dataset contains one report item. Its decision rows use these closed codes:

Score rule	Outcome	Reason code
score >= acceptThreshold	accepted	at_or_above_accept_threshold
reviewThreshold <= score < acceptThreshold	review	at_or_above_review_below_accept_threshold
score < reviewThreshold	rejected	below_review_threshold

The result also includes counts, inputDigest, rowsDigest, one digest for each ID list, and resultDigest. OUTPUT records whether Dataset delivery and the result-found charge were confirmed. Validation, budget, counter, and uncertain delivery failures remain non-success receipts and never retry a paid Dataset write.

Related Actors

Need to prepare, aggregate, and act on signal data?

Actor	What it does
US Brand Signal Metrics Aggregator	Summarizes already collected signal rows into deterministic counts and metrics.
US Brand Evidence Snapshot Diff	Compares two submitted evidence snapshots without fetching either one.
US Brand Action Queue	Turns selected signal rows into a bounded next-action proposal.
US Brand Offer Evidence Normalizer	Normalizes buyer-supplied offer evidence before policy gating.
Website Content Crawler	Collects website content upstream when a separate source-collection job is appropriate.

These five links returned HTTP 200 on a read-only verification on 2026-08-05. The linked Actors are separate products; this gate does not call them.

Limitations

• A score is an input value, not a truth probability. An accepted ID is not a verified lead, event, brand, person, or company.
• IDs are opaque and must be unique. Duplicate IDs are rejected rather than silently collapsed.
• Scores must be integers from 1 to 100; decimals, strings, missing fields, and extra fields fail closed.
• The maximum batch is 100 rows, and the report is intentionally compact.
• There is no live lookup, evidence validation, source freshness check, identity resolution, network access, proxy, BYOK, LLM, child Actor, state, or clock use.

FAQ

Can I use equal accept and review thresholds? No. The ordering is deliberately strict so every score has exactly one outcome: reviewThreshold < acceptThreshold.

Does accepted mean true or verified? No. It means only that the submitted integer met the buyer's stated threshold.

What happens at a boundary? Accept is inclusive. Review is inclusive at reviewThreshold and exclusive at acceptThreshold. Rejected is strictly below reviewThreshold.

Can I submit brand names or person names? They are outside this contract. Submit an opaque buyer ID; the Actor performs no identity inference.

What happens if a Dataset write or charge is uncertain? The Actor makes no second Dataset push. OUTPUT reports delivery_unknown, result_uncharged, or result_charge_delta rather than claiming success.

How do I report a problem? Use the issue form on the Actor page and include the schema version, error code, and a redacted input shape. Do not include secrets.

Built by zinin.