Social Content Change & Opportunity Monitor

Compare provided structured snapshots and emit deterministic change and opportunity rows. This Actor performs no autonomous scraping and does not acquire data from websites by itself; it only processes records supplied by the user or by an authorized upstream dataset.

What It Does

It joins current and prior snapshots by configurable identity fields, classifies added, removed, changed, unchanged, and stale entities, computes metric deltas, and emits explainable rows for items that should be reviewed.

Who It Is For

Analysts, marketing operators, agencies, and data teams that already have structured public-data exports and need repeatable change detection instead of manual spreadsheet comparison.

Input

Provide currentRecords and optional priorRecords. Configure idFields, metricFields, timestampField, growthAlertPct, and staleDays to match your data. If priorRecords is empty, the run emits first-snapshot rows without pretending a trend exists.

Output

Each dataset row includes entity identity, change type, deterministic score, metric deltas, changed fields, timestamps, source indexes, and ScoreExplanation.

Example

{
  "priorRecords": [
    { "url": "https://example.test/a", "likes": 100, "capturedAt": "2026-09-01T00:00:00Z" }
  ],
  "currentRecords": [
    { "url": "https://example.test/a", "likes": 150, "capturedAt": "2026-09-10T00:00:00Z" }
  ],
  "idFields": ["url"],
  "metricFields": ["likes"],
  "growthAlertPct": 25
}

How It Works

The algorithm normalizes identity values, removes duplicates by entity key, compares current and previous records, computes absolute and relative metric deltas, applies user-configurable thresholds, and ranks rows by score. Added and removed entities receive explicit classifications. Stale rows are based on timestamp age relative to the newest current timestamp in the run.

Priority and recommended action are derived from explicit score thresholds: score 70 or higher is treated as a high-priority review action, score 40-69 is a medium-priority review action, and lower scores are monitor-only unless the row is needed for added, removed, or stale reporting. Every score factor, action threshold, and rule trigger is included in ScoreExplanation.

Pricing

PAY_PER_EVENT event change_record_emitted is charged once per emitted change row. Invalid rows, duplicates, empty comparisons, and rows skipped by charge limits are not billable.

Limitations

This is a comparison and analysis Actor, not a data acquisition Actor. It does not verify whether an upstream dataset was collected correctly. Thresholds are configurable because useful alert sensitivity depends on the buyer workflow and dataset scale.

Integration

Run it directly with JSON input, from an Apify task, or downstream from another Actor that already produces structured records. Results are written to the default dataset and the run summary is written to the OUTPUT key-value store record.

Methodology

The executable spec generated by Agentic Implementation Planner V1 is stored with the Actor as .actor/executable_product_spec.json. The implementation follows that spec with deterministic tests for added, removed, changed, stale, duplicate, and malformed records.