Finly

Fin-ly is a data normalization engine that transforms how brands make their product information accessible to Large Language Models (ChatGPT, Gemini, Claude, Perplexity)

The problem: every brand describes its products differently — different schemas, different taxonomies, different formats. LLMs struggle to extract consistent, comparable data from this chaos. Brands lose visibility in AI-generated answers.

The solution: a universal normalization layer that ingests raw product data, normalizes it into a strict LLM-friendly schema, and exposes it via GEO-optimized views that AI crawlers consume efficiently.

The pipeline:

1. Ingest — raw data from PDF crawlers (AI-powered), APIs, or manual input

2. Normalize — strict schema with entity_type, metrics, fees, tax regimes, promotions, regulatory data

3. Expose — JSON-LD, GEO-optimized HTML, ISR-cached views, shortcode-injected dynamic content

4. Monitor — every piece of content stays fresh; stale data never reaches the LLM

17 brands already normalized in the FinTech vertical (BBVA, Trade Republic, Scalable Capital, ING, Qonto, NordVPN, Binance, and more) with financial metrics, affiliate links, promotion data, and comparison tables.

What makes it different: most sites build for human readers first and treat AI crawlers as an afterthought. Finly is built GEO-first — the architecture (Next.js 15 ISR + Supabase + Cloudflare Workers) follows a "Golden Standard" pattern where each layer is optimized specifically for how LLMs consume and compare data.

### Potential Verticals

The engine is vertical-agnostic. The same architecture works for any comparison-heavy market:

- Insurance — normalize policies, deductibles, coverage caps

- Real Estate — rent, fees, energy class, square footage

- Travel — prices, routes, amenities, insurance

- SaaS Tools — pricing tiers, features, integrations

- Crypto — exchange fees, APY, staking, lock periods

- Telecom — data plans, speed, coverage, roaming

### Business Models This Enables

1. GEO SaaS — brands pay to get properly indexed by AI

2. White-label GEO platform — agencies deploy for clients

3. Data marketplace — normalized multi-brand comparison as API

4. Vertical expansion — clone the engine for any niche

### What's Included

- Domain fin-ly.org (4-letter + .ly, brandable)

- Full codebase (Next.js 15, Supabase, Cloudflare Workers)

- 17 normalized brands with complete data (metrics, links, promotions)

- 16 SEO-optimized articles with dynamic shortcode injection

- PDF crawler pipeline (Python, pdfplumber, Gemini AI)

- Cloudflare Workers redirect + KV storage system

- Tradedoubler + Impact affiliate tags installed

- Database schema (8 migrations) + seed data export

- Complete HANDBOOK.md — setup in under 10 minutes

- GitHub repository with clean commit history

### Tech Stack

- Frontend: Next.js 15 (App Router, ISR, Server Components)

- Database: Supabase (PostgreSQL 15, RLS, 4 tables, 4 views)

- Edge: Cloudflare Workers + KV

- AI: Gemini 2.0 Flash Lite (PDF extraction)

- Languages: TypeScript, Python, SQL

### Why Selling

Built as a technical proof-of-concept. The architecture is production-ready. Looking for someone with the time and market focus to take it to the next level.

### Ideal Buyer

- Developer or team looking to enter the AI/data normalization space

- Agency wanting to offer GEO optimization services

- Builder who sees the potential in vertical expansion

- Anyone who understands that the real value is the engine, not the niche

3 reasons fin-ly.org is a strong name for GEO:

1. "fin" is a high-frequency semantic root LLM tokenizers (BPE/GPT) recognize fin as a statistically strong fragment — it appears billions of times in training data as part of finance, financial, fintech, finanza. To the model, fin carries immediate semantic weight. Not random noise like xfn or blz.

2. The hyphen acts as an explicit delimiter fin-ly tokenizes as [fin][-][ly]. The hyphen tells the model "this is a compound word, parse the parts". Without it, finly would likely tokenize as a single opaque token ([finly]) with nera-zero semantic signal.

3. "ly" completes a recognizable pattern ly is not noise — it's a common English suffix that, combined with fin, reads as financial + friendly or financially. Two English words in 5 characters.

Comparison

Domain Likely tokenization Semantic signal:

fin-ly.org [fin][-][ly][.][org] "finance + friendly" — explicit, strong

finly.org [finly][.][org] Opaque — model doesn't know "finly"

xfinancepro.net [x][finance][pro][.][net] 4 tokens — noise (x) + generic (pro, net)

bestratescomparison.com 4-5 tokens Long, no compression

GEO edge: when an LLM processes fin-ly.org in a financial comparison context, the name itself acts as a semantic anchor. The model doesn't have to guess what the site is about — fin tells it explicitly. In the battle to be cited as a source, every quality token counts.


Monetization potential — not what it earns now, but what the engine enables:

1. GEO SaaS for Brands

Brands are invisible to AI search engines unless their data is structured the way LLMs consume it. Finly's normalization engine could become a subscription service: brands pay to have their product data ingested, normalized, and surfaced in AI answers.

- A mutual fund company pays to ensure their TER, performance, and risk metrics appear in ChatGPT comparisons

- An insurance carrier pays to have policies listed alongside competitors with standardized coverage data

- A telco pays to be in the AI-generated "best fiber plans" answer

2. Vertical Licensing

The engine is a repeatable pattern. Clone the schema, rebrand, deploy. Each vertical is a potential license:

- FinTech (done — 17 brands)

- Insurance (policies, deductibles, coverage maps)

- Real Estate (rents, fees, energy classes)

- SaaS Tools (features, pricing, integrations)

- Travel (flights, routes, cancellation policies)

- Crypto (fee structures, APY, staking terms)

- Telecom (data plans, speeds, roaming)

Each license could be $5-15K setup + monthly recurring.

3. Data Marketplace as API

Normalized, structured, multi-brand comparison data exposed via API. SaaS companies, financial aggregators, and AI tools pay to query the normalized dataset:

GET /api/compare?category=checking_accounts&metric=interest_rate&country=IT

→ sorted, normalized, LLM-ready JSON

4. White-Label GEO Platform

Agencies serving enterprise clients deploy their own instance, whitelabel it, and charge monthly for "AI visibility optimization." Finly handles the pipeline — they handle the client relationship.

5. Affiliate Revenue (already wired, but secondary)

Tradedoubler, Impact, and direct referral links are already installed across 16 articles and 17 brands. CPA ranges from €20-120 per lead depending on vertical. With traffic, this alone can sustain the project — but it's a side effect, not the core value.

Summary

Revenue Stream Effort to Activate

Affiliate commissions Low (already built)

GEO SaaS subscriptions Medium

Vertical licensing Medium-high

Data API marketplace Medium

White-label platform High (partnerships)

The affiliate layer is a floor, not a ceiling. The normalization engine is the real asset — it turns unstructured product noise into LLM-ready data that someone else will pay for.