Google Play Credit App Reviews Analysis

What does google-play-credit-app-reviews-analysis do?

This Actor analyzes negative reviews (rating 1-2) from a target Google Play credit app page and extracts core lending signals from user complaints. It calls the Apify Store Actor webdatalabs/google-play-reviews-scraper to collect reviews, then uses OpenRouter to call a large language model (fixed model: qwen/qwen3.7-plus) to infer:

• Loan disbursed amount (actual arrival amount)
• Repayment amount
• Repayment term
• Fee rate = (repayment amount - arrival amount) / repayment amount

It outputs per-review structured records and one summary record with amount range, amount median, repayment term distribution, and fee rate median.

Why use this Actor?

• Understand user-reported loan economics from complaint-heavy reviews.
• Estimate real user burden from review text, not only app metadata.
• Identify suspicious repayment conditions and abnormal fee rates quickly.
• Build data feeds for risk dashboards, compliance checks, and product benchmarking.

How to use this Actor

1. Open the Actor input form.
2. Set Google Play app URL in startUrls.
3. Set scraping settings (maxReviews, sortBy, proxyConfiguration).
4. Optionally adjust confidenceThreshold and outlier-statistics behavior.
5. Run the Actor.
6. Read dataset output for review-level rows and one recordType = summary row.

Environment Setup

Before running this Actor, configure these environment variables:

Local testing (PowerShell):

$env:APIFY_TOKEN = "your-apify-api-token"
$env:OPENROUTER_API_KEY = "your-openrouter-api-key"
apify run

Permanent configuration (Windows):

[Environment]::SetEnvironmentVariable("APIFY_TOKEN", "your-apify-api-token", "User")
[Environment]::SetEnvironmentVariable("OPENROUTER_API_KEY", "your-openrouter-api-key", "User")

On Apify Platform: Add OPENROUTER_API_KEY to the Actor's Secrets in the Apify Console. APIFY_TOKEN is automatically provided by the platform.

Get your tokens at:

• Apify Console → Account → API tokens
• OpenRouter Dashboard → Dashboard → API keys

Input

This Actor accepts the same scraping controls as the upstream scraper:

• startUrls (required): Google Play app URLs.
• maxReviews: max number of reviews requested from upstream actor.
• sortBy: sort mode passed through to upstream actor.
• proxyConfiguration: proxy config passed through to upstream actor.

Analysis controls:

• confidenceThreshold (default 0.6): review extraction confidence threshold.
• includeOutliersInStats (default false): whether fee-rate outliers (<10% or >80%) are included in core fee median.

Internally forced values when calling upstream actor:

• rating = NEGATIVE
• language = en

Output

Output is pushed to default dataset.

• Multiple rows with recordType = review.
• Exactly one row with recordType = summary.

Simplified example:

[
    {
        "recordType": "review",
        "appId": "com.example.loan",
        "reviewRating": 1,
        "arrivalAmount": 7000,
        "arrivalCurrency": "KES",
        "repaymentAmount": 9100,
        "repaymentTermNormalized": "7",
        "feeRate": 0.2308,
        "feeRateOutlier": false,
        "extractionConfidence": 0.82,
        "extractionQuality": "HIGH"
    },
    {
        "recordType": "summary",
        "appId": "com.example.loan",
        "processedReviewCount": 78,
        "qualityStats": {
            "highQualityCount": 28,
            "mediumQualityCount": 34,
            "lowQualityCount": 16
        },
        "arrivalAmountByCurrency": {
            "KES": { "min": 1500, "max": 12000, "median": 6000, "count": 74 }
        },
        "repaymentTermDistribution": { "7": 44, "15": 17, "30": 9, "X*3": 8 },
        "feeRateMedian": 0.26
    }
]

You can download the dataset in various formats such as JSON, HTML, CSV, or Excel.

Data table

Field	Description
arrivalAmount	Disbursed amount detected from review text
repaymentAmount	Repayment amount detected from review text
repaymentTermNormalized	Normalized term (e.g. 7, 15, 30, 90, 7N, XN)
feeRate	Computed or extracted fee rate ratio
feeRateOutlier	True if fee rate < 0.1 or > 0.8
extractionConfidence	LLM extraction confidence score

Pricing / Cost estimation

How much does it cost to analyze one app?

• Upstream scraping cost depends on review volume and proxy usage.
• LLM cost scales with number of negative reviews and review text length.
• Start with smaller maxReviews (e.g. 100) to validate data quality first.

Tips and advanced options

• Keep confidenceThreshold around 0.3 to retain more records, then filter by extractionQuality (HIGH/MEDIUM/LOW) in downstream analysis.
• Increase threshold if your workflow is strict and low-noise.
• Keep includeOutliersInStats = false for stable central tendency metrics.
• If many records are filtered out, check language quality and app region conventions.

FAQ, disclaimers, and support

• Review text is user-generated and may be incomplete, sarcastic, or contradictory.
• Multi-language and slang can reduce extraction accuracy.
• Currency detection follows text clues and may return UNKNOWN.
• This Actor provides analytical inference, not legal or financial advice.

If you want a custom extraction schema or country-specific term/currency handling, open an issue in your repository and extend the extraction prompt and post-processing rules.