PA Report Workflow

PA 报告更新工作流

Overview

Weekly pipeline that updates the combined PA HTML report page with fresh data from AppsFlyer CSV exports.

Supported CSV File Types

| File Type | PA Category | |-----------|-------------| | blocked-installs | Install PA | | detection | Install/Event PA (check Event Name) | | blocked-in-app-events | Event PA | | fraud-post-inapps | Event PA |

Key Configuration

App Name Mapping:

APP_NAMES = {
    'idXXXXXXXX': 'App A iOS',
    'com.example.appa': 'App A Android',
    'idXXXXXXXX': 'App B iOS',
    'com.example.appb': 'App B Android',
}

Cross-Period Split:

# Some apps span Week 2 and Week 3
SPLIT_DATE = "YYYY-MM-DD"  # Configurable split point
# Install Time < SPLIT_DATE -> Week 2
# Install Time >= SPLIT_DATE -> Week 3

Execution Flow

1. Confirm CSV files downloaded to Downloads folder
2. Update WEEK_FILES config to match actual filenames
3. Run generate_week_data.py:
   - Read each CSV file
   - Parse per (channel, app_id, os)
   - Aggregate: install_pa, event_pa, reasons
   - Output: week2_final.js + week3_final.js (JSON arrays)
4. Run inject_data.py:
   - Read generated JS files
   - Inject into pa_report_combined.html
   - Replace var week2Data / week3Data declarations
5. Validate output JSON is well-formed

Output Data Schema (per record)

{
  "channel": "example_channel",
  "app_id": "com.example.app",
  "os": "Android",
  "install_pa": 55,
  "event_pa": 22468,
  "install_reasons": ["bots/bayesian_network"],
  "event_reasons": ["bots/behavioral_anomalies"],
  "pid": "",
  "pid_details": [
    {"pid": "example_pid", "install_pa": 0, "event_pa": 8941}
  ]
}

Known Pitfalls

1. reasons must use list, not Counter: Counter has no .append() method
2. inject_data.py handles single-line empty arrays: var week3Data = []; needs special parsing
3. CSV filename dates may vary: Always verify filenames before running

Scripts

The following bundled scripts support this skill:

| Script | Purpose | |--------|---------| | pa_report_workflow.py | Executable script |

Dependencies

pandas >= 1.5.0
# No external dependencies for injection step (string manipulation only)