DropFix

Retention signal -> exact moment -> edit decision.

The problem

YouTube Studio can show a creator where viewers leave. It rarely tells an editor what to change.

A retention drop might come from a slow setup, a missing payoff, a confusing transition, or something outside the video. Today, creators have to manually compare a graph with a transcript, guess at the cause, and translate that guess into an editing brief.

The solution

DropFix turns retention evidence into an editor-ready decision.

A creator imports a transcript and, when available, a YouTube Studio CSV export. DropFix detects meaningful retention changes, aligns each signal with the matching transcript moment, and produces a prioritized recommendation with:

• Observed: the measured behavior change
• Compared: the available baseline context
• Interpretation: a clearly labeled likely explanation
• Recommended action: what to cut, move, rewrite, preserve, or test
• Confidence: calibrated to the evidence, not artificial certainty
• Validation: the metric or future test that would show whether the change worked

Editors can accept, mark for rewrite, or dismiss each recommendation, then export those choices as a clean editor brief.

Why this is different

Most creator analytics tools stop at the graph. Most AI tools start with generic advice.

DropFix connects the whole chain:

metric -> moment -> explanation -> edit

Instead of "improve the hook," an editor gets an exact retention interval, the linked transcript passage, a concrete edit, why the recommendation is credible, and what to measure next.

Built to be trustworthy

DropFix uses deterministic retention-event detection and a structured recommendation schema rather than pretending every graph movement has a certain cause.

Its visible 40-point quality check evaluates evidence quality, specificity, actionability, goal alignment, confidence calibration, expected usefulness, novelty, and validation quality. That score measures the quality of a recommendation, not predicted YouTube performance or hackathon placement.

The live demo uses clearly labeled synthetic data so anyone can test the full workflow without an account. The same import path accepts creator-provided YouTube Studio CSV exports for creator-specific retention evidence. DropFix does not scrape YouTube or claim direct account integration.

Built with

HTML, CSS, vanilla JavaScript, browser local storage, client-side transcript and CSV parsing, and GitHub Pages.

Creator impact

DropFix saves the repetitive work between "I can see a drop" and "here is the exact change my editor should make." It gives creators less analytics interpretation, fewer vague revision notes, and more time making better videos.

Built by Ezra Westover as a solo project.

Built With

• css
• csv-parsing
• github
• html
• localstorage
• vanilla-javascript