Launched this week
Kehai
Cookieless web analytics with a forecast that grades itself
2 followers
Cookieless web analytics with a forecast that grades itself
2 followers
Kehai is cookieless web analytics hosted in the EU. Reports are counted from recorded events without sampling, and imported or modeled numbers keep their own labels. The forecast tests itself on your history and shows its error next to the answer. Pacing tells you mid-month whether a target is still in reach. Sales and refunds come from the shop itself, AI referrals and bots get their own reports, and a read-only MCP server lets your assistant ask.
Hey Product Hunt! 👋
I'm Mikołaj, the maker of Kehai. Today I'm launching it here: cookieless web analytics with a forecast that grades itself, run from Warsaw and hosted in the EU.
What can you do with Kehai?
📈 Forecast a goal's completions or its net revenue, with the forecast's own error next to the answer
🎯 Set a target for the month or quarter and see the run rate you need beside the one you have
💶 Get revenue from the shop itself: confirmed sales and refunds from Stripe, WooCommerce, PrestaShop, Wix, or Squarespace
🤖 See which AI assistants send you people, and every excluded bot with the rule that caught it
🔌 Let Claude, Cursor, or VS Code read your reports through a read-only MCP server
📦 Bring your history from GA4, Plausible, or Umami, with its origin on every day
🪶 Run all of it from a 2.5 kB script, with no cookies and nothing in browser storage
✨ How does the forecast work?
It forecasts one outcome: a goal's completions or its net revenue. Before it answers, it tests itself on your own history against the simplest baseline there is, repeating last week. Next to the forecast you see its skill and how often its 80% band actually held. With too little history it refuses and says what's missing.
The same rule runs through the whole product: counted, imported, and modeled numbers never share a column, and nothing is sampled.
🎯 Motivation
I've worked in digital marketing for 15 years, and today's popular analytics tools are why I built Kehai. Their scripts are heavy, and I trusted their reports less every year: sampling, thresholds, and modeled numbers mixed in with the counted ones. I wanted analytics where every number says how it was made: counted, imported, or modeled.
💬 Good to know before you click
- It's paid and built for businesses, from €49 a month. Pacing, MCP, and imports are in every plan. The forecast starts at Growth, revenue reports at Commerce.
- Every plan starts with 7 days free, and Checkout asks for a card.
- For Product Hunt: PRODUCTHUNT30 takes 30% off your first month on a monthly plan, until October 17.
- There's no session replay, no heatmaps, and no user profiles. Use Hotjar for that. We won't build it.
- It's hosted only. Not open source.
👀 Look around without an account
The demo opens on two years of generated shop data, with the forecast, pacing, and revenue full: https://kehai.io/sign-in?demo=1
Thanks, everyone! I'd love to hear what you'd check before trusting a forecast, and where the counting is unclear.
I'm here all day.
Best,
Mikołaj