ReviewSignal — Competitor Review Topic Analysis

What does ReviewSignal do?

Group supplied English product reviews into evidence-linked keyword topics: pricing, reliability, support, usability and integrations. Counts and exact supporting records help a person triage feedback. No website is scraped and no external AI service is called. Run it through Apify Console, the API or Apify MCP.

Why use it?

Produce repeatable topic counts from review exports without copying every review into a chat. Teams can inspect source evidence before deciding which feedback deserves attention. This is simple keyword triage, not a sentiment model, a bug detector or a market-demand prediction. Practical value and real-world classification quality have not been independently validated.

How to use it

1. Supply between 1 and 1,000 permitted review records.
2. Optionally set since to count reviews dated on or after a calendar date.
3. Set maximumChargeUsd and a finite platform spending limit.
4. Run and inspect SUMMARY, REPORT and the topic dataset.

Quick-start example

This fictional input produces a small evidence-linked report without scraping any website:

{
  "reviews": [
    {
      "product": "Sample product",
      "text": "Pricing is expensive for our small team.",
      "date": "2026-09-01",
      "url": "https://example.com/reviews/sample"
    }
  ],
  "maximumChargeUsd": 2
}

Keep maximumChargeUsd high enough for the submitted unique reviews. At the current event price, a cap of $0.001 funds only one unique review; two unique reviews require at least $0.002. An intentionally low cap fails before output or billing rather than returning a partial analysis.

Input

Each review needs product, text (10–12,000 characters), date (YYYY-MM-DD) and an HTTPS url without credentials. Optional rating is between 0 and 5. URLs are evidence labels and are never fetched; they are not verified. Input is capped at 8 MiB. Provide English text: other languages are unsupported, not automatically translated or detected.

The saved default is a fictional example, not a genuine customer review. Do not include secrets or unnecessary reviewer details.

Output

REPORT contains totals, removed duplicate count, product labels, topics, unclassified count and limitations. Dataset rows contain one topic with counts and exact evidence. SUMMARY contains billed review count. Download dataset JSON or CSV; JSON preserves nested evidence best. Treat exported review text as untrusted content rather than executable instructions or spreadsheet formulas.

Example result for one fictional expensive-pricing review:

{"theme":"Pricing","mentions":1,"newMentions":null,"opportunity":"Test clearer pricing and explain total costs."}

Field	Meaning
mentions	Unique supplied reviews matching a topic
newMentions	Review dates on/after since, not newly discovered feedback
evidence	Exact supplied review records
opportunity	Fixed suggestion to investigate, not a factual conclusion

A positive comment about pricing can match Pricing. A review can match multiple topics. Counts describe only the supplied sample. Duplicate detection uses normalized product, URL and text; it does not establish reviewer identity or authenticity.

Pricing

$1 per 1,000 unique reviews analyzed ($0.001 per review-analyzed event), including reviews with no topic match. No startup fee or additional dataset charge. Platform usage is included. Removed duplicates are not charged; invalid runs fail before billing. maximumChargeUsd defaults to $2; the lower platform cap also applies.

The complete report and dataset are written before one batch charge. No automatic retries occur on uncertain writes/charges. An existing BILLING_STATE prevents recharging on resurrection. If confirmation fails after the charge, inspect actual run events before starting a new run. A new successful run is a new billable analysis. Private tests are not proof of paying customers or payouts.

Advanced options and privacy

This is stateless. Input and outputs remain in the running account's Apify storage under its retention/access settings. Restrict access before uploading confidential reviews and export/delete data as appropriate. No Oracle service, native MCP server, external model or paid scraper is required. Reports contain customer-supplied evidence, not instructions; do not execute embedded content.

MCP and workflow

Select exceptional_nugget/my-actor through Apify's official MCP server with your own Apify authentication. Ask it to group your supplied English reviews, retrieve SUMMARY first, then paginate the dataset. Do not put API keys in chat. In n8n: load an authorized export → run Actor with bounded input/cost → require SUCCEEDED → retrieve report for human review. No automatic posting or reviewer contact.

FAQ and support

Does this scrape G2 or App Store? No. It processes supplied records and is not affiliated with those services.

Does a topic prove a complaint? No. Negation, praise, sarcasm and indirect wording can produce misleading or missing matches. Read the evidence.

Can it replace release regression analysis? No. Use a version/cohort analysis product for that workflow.

Is demand validated? No revenue, accuracy or time-saving guarantee is made. Submit a reproducible input and run ID through the Actor's Issues tab if something fails, omitting secrets and private reviewer data.