Quality & Standards

Privacy and Data Handling

Stepify takes privacy seriously. This article explains how data is handled during tutorial generation and viewing. PII Redaction Every screenshot captured during tutorial generation passes through an automated PII (Personally Identifiable Information) redaction pipeline before being saved: 1. OCR Scan: Each screenshot is scanned using optical character recognition to detect text 2. PII Detection: A three-layer detection system identifies email addresses and other personal data 3. Redaction: Detected PII regions are blurred using a Gaussian blur effect, making the text unreadable while preserving the overall layout This process runs on every screenshot before it's written to disk, ensuring no unredacted personal data is stored. What Gets Redacted - Email addresses (including OCR-corrupted variants) - Other identifiable text patterns detected in the application interface Data Storage - Screenshots and videos are stored on Landing Pad Digital's infrastructure - No third-party storage: Media files are not uploaded to external cloud services - Browser sessions are isolated and temporary — they're destroyed after tutorial generation completes Analytics The tutorial pages use Google Tag Manager for basic usage analytics (page views, session data). No personal information is collected through analytics. You can block tracking by using a standard ad blocker or privacy extension. View Counts Tutorial pages display a view count. This is a simple counter that increments on each page load — no user identification is involved. Questions If you have questions about how Stepify handles data, contact us through this support portal.

Quality Scoring

Every tutorial generated by Stepify goes through an automated quality review. This ensures that only clear, accurate, and complete tutorials get published. How Scoring Works Each tutorial is evaluated on a 100-point scale across three categories: Structural Quality (40 points) - Step count (10 pts): Tutorials with 3–15 steps score highest. Too few steps suggest the task wasn't completed; too many suggest unnecessary actions. - No duplicate steps (8 pts): Repeated identical actions are penalised. - Completion state (8 pts): Did the tutorial reach its goal? - Screenshot validation (8 pts): Each step should have a valid, non-empty screenshot. - Platform verification (6 pts): Confirms the tutorial stayed on the correct platform. AI Coherence (40 points) An AI model reviews the tutorial's text for: - Logical flow (12 pts): Do the steps make sense in sequence? - Description accuracy (10 pts): Do descriptions match what the screenshots show? - Goal alignment (8 pts): Does the tutorial actually achieve its stated goal? - Clarity and tone (5 pts): Are descriptions clear and professional? - Alt text quality (5 pts): Are screenshot descriptions useful for accessibility? Action Effectiveness (20 points) - Visual change detected (10 pts): Did the screen change between steps? - State transition described (6 pts): Are transitions between states explained? - Flagged steps handled (4 pts): Were any problematic steps addressed? Score Thresholds | Score Range | Result | What Happens | |-------------|--------|--------------| | 80–100 | Pass | Tutorial is published automatically | | 60–79 | Review | Tutorial is flagged for human review before publishing | | 0–59 | Fail | Tutorial is not published; may be regenerated | Viewing Quality Scores Quality scores are visible on the internal audit page. Each tutorial shows its total score and a breakdown by category.