Conversion on /checkout fell 18%
Compared with the same Tuesday over the last four weeks.
- What
- Product analytics suite
- What we did
- Strategy, design, development and AI
- Live at
- eyes.dev
We wanted to know how our software was actually being used
We build software for a living, for ourselves and for clients. After every release we wanted the same answers: is anyone using it, where do people drop off, and what should we fix next? The analytics tools we tried mostly gave us dashboards. Plenty of charts, but the digging was still on us.
So we built Eyes. Not a tool you lose hours in, but one that keeps watch, tells you what changed, and answers your questions with your own numbers. Answers instead of dashboards, minutes instead of hours.
Ask a question, get an answer
Ask what's going on in plain English. Eyes pulls the numbers first, then backs up its answer with charts and funnels right there in the conversation. The model never writes its own queries: it picks from a fixed set of tools, and every value is checked on the server. That's how Eyes avoids making numbers up.
Askhttps://yourproduct.comNew chat
Eyes keeps watching, even when you don't
Every fifteen minutes, Eyes checks your numbers against what's normal: the same day of the week, over the last four weeks. When conversion drops, a page slows down or traffic goes quiet, you get an alert with the context attached.
You don't have to set it all up first. Eyes learns how your site behaves and suggests what's worth keeping an eye on.
/pricing is loading noticeably slower
LCP went from 1.9 to 3.4 seconds after yesterday’s release.
New traffic from news.ycombinator.com
Most of it lands on your blog and stays longer than average.
More drop-off at step 2 of your onboarding
Drop-off rose from 22% to 31% this week, mostly on mobile.
No events from your app for 45 minutes
This is usually a busy hour. Is the script still on the page?
Returning visitors are up
12% above a normal Wednesday, since the newsletter went out.
Everything you expect from analytics
Under Ask and the alerts sits a complete analytics platform. One script, and Eyes tracks pageviews, sources, engagement and Core Web Vitals. Funnels, retention and visitor profiles build on top of that, for your product and your marketing site at once.
Visitors, sessions, bounce rate and conversion at a glance, always compared with the previous period.
Your analytics, inside your AI tools
If you work with AI agents, the last thing you want is your data in yet another tab. Eyes runs an MCP server, so you can ask the same questions from Claude, ChatGPT, Cursor or any other MCP client. Your agent only ever gets aggregated numbers back, never individual visitors.
claude · ~/webshop
>
⏺ eyes · query_analytics (metric: conversion_rate, dimension: device, compare: previous_period)
⎿ mobile 2.1% (−0.8 pt) · desktop 3.9% (+0.1 pt) · tablet 3.2% (±0)
⏺ eyes · query_funnel (funnel: checkout, device: mobile)
⎿ biggest drop-off: shipping → payment, 41% leave (was 24%)
⏺ The drop is all mobile, and almost all of it sits between shipping and payment. That step changed in Tuesday’s release; desktop hasn’t moved. I’d start with the payment form on small screens.
Counting visitors without following them
Eyes doesn't need a cookie banner. Every visitor gets an anonymous ID that lasts a day, and IP addresses never reach the database. Because the ID changes daily and differs per site, nobody can follow a visitor across days or across websites.
When someone does give consent, Eyes remembers them with its own cookie. Only then do retention and visitor profiles work for that person.
EyesSee what’s happening in your product
Ask a question in plain words and Eyes answers it from your own data. When something shifts, it tells you before you think to look.
