Review Response Prioritizer by Location

Rank which locations and complaint topics need attention first, using review data you already have. The Actor does not scrape Google Maps or any other site, makes no network requests, and never reads or outputs reviewer names or profile data.

What it does

1. Reads your review records (location, rating 1–5, date, optional text).
2. Compares a recent window (default 30 days) with the equally long period before it.
3. Detects complaint topics in low-rated reviews with transparent English keyword rules (plus optional custom topics).
4. Outputs two ranked record types in one dataset:
   • location records: priority score, recent low ratings, rating change versus the prior period, recurring issues, confidence label.
   • topic records: topics ranked across all locations, with locations affected and recurrence counts.
5. Saves a SUMMARY record with window dates, input statistics and limitations.

Priority score (0–100)

min(100, 12 × recency-weighted recent low ratings + 8 × recurring topics + 10 × rating drop)

• Recency weight falls linearly from 1.0 (as-of day) to 0.5 (oldest day of the window).
• A recurring topic appears in at least minRecurrence recent low-rated reviews at that location.
• Rating drop counts only when both windows have at least 3 reviews.

Sample input

{
  "reviews": [
    {"locationName": "Harbor Cafe", "rating": 1, "publishedAt": "2026-10-20", "text": "Long wait and the staff were rude."},
    {"locationName": "Harbor Cafe", "rating": 2, "publishedAt": "2026-10-18", "text": "Wrong order and slow service."},
    {"locationName": "Central Gym", "rating": 1, "publishedAt": "2026-10-22", "text": "Dirty showers and a bad smell."},
    {"locationName": "Central Gym", "rating": 2, "publishedAt": "2026-10-15", "text": "Locker room is filthy."}
  ],
  "recentDays": 30,
  "lowRatingMax": 2
}

Accepted field aliases: location (locationName, location, placeName, place), id (locationId, placeId), rating (rating, stars), text (text, review, comment), date (publishedAt, date, reviewDate). Dates must be ISO format.

Input options

Field	Default	Meaning
reviews	required	Your review records
recentDays	30	Recent window length (7–180)
lowRatingMax	2	Ratings at or below count as low
asOfDate	newest review date	Reference day for windows
minRecurrence	2	Low-rated mentions needed to call a topic recurring
useDefaultTopics	true	Use built-in keyword topics
customTopics	none	Your own topic → keyword lists
maxLocations	50	Cap on ranked locations
maxReviews	20000	Cap on reviews processed

Limitations and sample-size caveats

• Only the data you supply is analysed. Missing or unrepresentative exports produce misleading rankings.
• Keyword topic detection is simple: it misses paraphrases, sarcasm and other languages, and can over-match. Treat topics as leads and read the original reviews.
• With fewer than 5 recent reviews at a location, confidence is low; 5–14 is medium; 15 or more is higher. Even "higher" is not statistical proof.
• The score is a relative triage aid, not a measure of true quality, and is not calibrated to any business outcome.
• Records with missing location, invalid rating, or invalid date are skipped and counted in SUMMARY. Reviews older than two windows or after the as-of date are not scored.
• The Actor does not detect fake or duplicate reviews.

Privacy

Reviewer names, profile links and photos are ignored. Review text is used only for topic matching and is not written to the output. Do not include data you have no right to process.

Billing

The code emits a location-priority-report event per ranked location after results are saved. Invalid input and failed runs end before any charge. The price, if any, is configured separately by the publisher.