FlyRank AI · Final Impact Project

Opportunity intelligence a reviewer can audit.

FlyRank Opportunity Intelligence Studio converts approved Google Search Console and GA4 exports into a ranked, explainable content-opportunity queue. It works locally, preserves incomplete evidence, and stops every recommendation at a human-review boundary.

Open live demo Review source Watch demo video

The problem

Dashboards show metrics. They do not explain what deserves attention first.

SEO and content teams can export thousands of GSC and GA4 rows, but prioritization still requires manual joining, data-quality judgment, and a defensible next action.

One safe input loop

Upload schema-compatible GSC and GA4 CSVs, or use the built-in synthetic demonstration. Files remain in the browser.

Explainable scoring

Every rank is supported by visible demand, engagement, position, and data-quality signals rather than a hidden model response.

Human control

The workflow can recommend a review, but it cannot edit a website, publish content, create tickets, or approve business intent.

How it works

A deterministic workflow with explicit handoffs.

ValidateIdentify the two schemas and reject missing or malformed required fields.

NormalizeClean landing-page paths before any aggregation or source alignment.

JoinUse a full outer join at landing-page level. Never invent a query-level GA4 key.

ScoreRank opportunities using transparent signals and documented recommendations.

ReviewExport the brief only after the reviewer sees warnings, rationale, and limitations.

Synthetic demonstration

The output is ranked, readable, and still marked for review.

The values below are public sample data, not client analytics.

Rank	Page and query	Score	Signals	Recommendation	Boundary
1	/services/automation workflow automation	74	3,500 impressions 1.6% CTR Position 7.8	Rewrite title and description, then verify query-to-page alignment.	Needs review
2	/case-studies/erp erp software	67	1,600 impressions 1.4% CTR Position 8.6	Strengthen the existing page around demonstrated search demand.	Needs review
3	/case-studies/monitor vps monitoring	57	900 impressions 1.6% CTR Position 6.9	Inspect title, evidence, and intent before changing content.	Needs review

Verification

Five cases test the parts most likely to produce a misleading report.

The public application and repository preserve the same quality boundary: no silent fixes, no invented data, and no autonomous publishing.

Successful new input
Missing required column
Blank anonymized query preserved
Unmatched landing pages preserved
Malformed numeric value rejected

Open automated test source →

Launch story

From a repeated analysis task to a bounded public product.

“The strongest result is not a clever score. It is a result another person can reproduce, inspect, challenge, and safely reject.”

I started with a recurring problem: GSC and GA4 exports contain useful evidence, but the path from rows to a defensible priority is slow and easy to overstate. The first version proved the fixed workflow. The personal-agent work added reusable documentation, tests, output formats, and clear guardrails.

The final capstone connects the product to a real personal brand: a permanent portfolio domain, a public repository, a working demo, a narrated run, explainability, launch hardening, and an honest limitations record. The result is intentionally bounded. It helps a person decide where to investigate; it does not pretend to replace business judgment.

Four-link reviewer path

Review the outcome without reconstructing the project.

01 · Use itLive browser demo

Run synthetic data, inspect scores, and export reports.

02 · See itEnd-to-end video

Watch the setup, tests, run, output, and limitations.

03 · Verify itPublic repository

Read the implementation, tests, README, and deployment workflow.

04 · Understand itCapstone documentation

Review the problem, decisions, evidence, evaluation, and launch record.