Why we're building Meelu in the open
Meelu won't ship for a few months. Here's what it is, who it's for, and why the pieces are becoming open source MCP servers before the app exists.
Blog
Notes from building Meelu in the open.
How to Use AI to Analyze CRM Data (HubSpot, Zoho, Salesforce, Pipedrive)
I loaded 3,000 closed leads from a B2B SaaS-style CRM export and let AI analyze them with real statistical tools — no spreadsheet wrestling. Here's which leads actually convert, what drives lead scores, and how to run the same AI lead scoring analysis on your own HubSpot or Salesforce export in 15 minutes.
Can ChatGPT or Claude Analyze CSV Files? Chatbot Uploads vs. a Real Analytics Engine
We asked the same 5 questions of a 3,000-row CRM dataset two ways: an AI reading the raw CSV with no tools, and an AI driving a real analytics engine. Here are the exact answers, the errors, and the token bill.
How to Analyze Email Marketing Data with AI (Mailchimp, Klaviyo, Any ESP)
I loaded 455 email campaigns and 5,000 subscribers into a local analytics MCP server and asked it plain-English questions: best send time, subject line length vs open rate, why people unsubscribe, and which campaigns actually earn revenue. Every number comes from a recorded statistical test.
How to Use AI to Analyze Google Analytics (GA4) and Search Console Data
I connected Claude to a real statistics engine and analyzed two years of GA4 data plus 600 Search Console queries for a personal-finance site. Here's how AI found the traffic drop, proved it wasn't noise, and surfaced 70,000 potential clicks — without hallucinating a single number.
How to Analyze Hotel Booking Data with ChatGPT or Claude
I asked Claude to analyze 4,500 hotel bookings: which reservations will cancel, which channels bleed revenue, and what next season looks like. Real numbers, honest confidence levels, and a step-by-step way to do it with your own PMS data.
How to Analyze Any CSV or Spreadsheet With AI (Without It Making Numbers Up)
How to analyze any CSV or spreadsheet with AI — without the hallucinated numbers. Why pasting data into a chatbot fails, how a local analytics engine fixes it, and nine real walkthroughs: Shopify orders, GA4 traffic, ad spend, CRM leads, SaaS churn, hotel bookings, and more.
How to Analyze Google Ads and Meta Ads Data with AI
I let Claude analyze six months of Google and Meta ads data — 3,496 rows — with a local analytics tool. It found creative fatigue, wasted budget, and a tracking outage, and it showed its work. Here's exactly what it found and how to do it with your own CSV export.
How to Analyze Real Estate Data with AI: What Really Drives Home Prices
I loaded 3,497 home sales into a local AI analytics tool and asked it what drives prices, why listings sit, and when to sell. Real model, real numbers, honest confidence levels — and how agents can run the same analysis on their own MLS export.
How to Analyze SaaS Churn Data with ChatGPT, Claude or Any AI
I asked Claude to analyze 3,000 SaaS accounts for churn drivers — retention cohorts, an AI churn model, and a causal check on onboarding. The most useful thing it did was admit what it couldn't prove. Real numbers, real charts, and a step-by-step guide to running the same analysis on your own subscription data.
How to Analyze Shopify Store Data with ChatGPT or Claude
I connected Claude to a free, local analytics tool and let it dig through 18 months of Shopify order data. Here's how AI found my best customers, product bundles, retention leaks, and a sales forecast it honestly refused to oversell.
How to Use AI to Analyze Business Expenses (QuickBooks, Xero, Any Export)
I used a local AI analytics server to analyze 4,692 business expense transactions: it found $18,000 in duplicate payments, a hidden 45% price hike, and $18,359 in anomalous invoices. Here's how to do it with your QuickBooks or Xero export.
Introducing meelu-analytics: deterministic data analysis over MCP
An open source MCP server that ingests your CSVs into DuckDB and answers analytical questions with honest confidence levels — including refusing to answer when the data can't support the question.