Predictive funnel analytics
Kassandra watches your signup-to-paid funnel. When checkout conversion slides from its usual 27–29% to 18.4% on a Tuesday afternoon, it says so in Slack, names the likely cause, and lists the accounts that were mid-checkout.
Oracle is $0. Prophet is $69 a month. Pantheon is $529.
The notice in the corner is what your team reads first. It lands in Slack before the standup does.
Funnels watched, quietly, for teams at
Yarrow & CoDunmoreAshgroveValewoodBrightwaterWhat it does
Each one used to be somebody’s Thursday afternoon. Now it happens at 14:20 on whichever day it needs to.
Kassandra compares every funnel step with the same hour on the same weekday over the previous four weeks. When checkout conversion sits more than five points under that baseline for 40 minutes, it posts to Slack with the step, the likely cause and the accounts that were mid-checkout.
Drag a chart onto the grid, point it at a connected table, and it draws itself in about two seconds. Resize it from the corner handle. The totals on the dashboard match the totals on the funnel page, because they read from the same place.
Choose PDF, Slack or both, and a time: Monday, 08:00 UTC. The digest opens with one sentence on what changed, then shows the three charts that explain it. Recipients can mute any section they don’t care about. Nobody can mute the sentence.
Kassandra projects each signup cohort six weeks out, then explains the forecast without a Greek letter anywhere. For example: the Oct 19 cohort will settle near 28%, because fewer of its members built a second dashboard in week one.
Setup
Pick Postgres, Stripe, Snowflake or HubSpot. Each connector asks for read-only credentials and pulls up to 30 days of history before you’ve finished your coffee.
Name the funnel steps, decide who hears about what (#growth-weekly, Priya N., Tomasz K.), and set a quiet hour so nobody gets paged at 03:00 about a Sunday.
Kassandra learns four weeks of weekday rhythm, then starts talking. Until then it only watches, and it tells you it is only watching.
Integrations
Alerts and reports land where the conversation is already happening, with the context attached.
Plans
Prices are per workspace, per month. Nobody’s count of dashboards is audited.
The one with a human analyst attached
For one product and one patient person.
$0 per month
For a team that has stopped checking the dashboard by hand.
$69 per month
For companies where a bad Tuesday has its own name.
$529 per month
In their words
A checkout step was timing out for one card network. Kassandra posted in Slack while I was still on my first coffee. The old dashboard would have shown it to us the following Monday.Priya N.Head of Growth, Dunmore
The Monday digest is one sentence and three charts. Our finance team reads it on the commute, which cannot be said for the spreadsheet it replaced.Hana B.Revenue Operations, Ashgrove
I asked why the October cohort was sagging and got a paragraph a board member could repeat. Nobody opened a notebook.Callum W.Founder, Brightwater
Oracle costs $0 and watches three data sources. Prophet costs $69 and watches fifteen. Pantheon costs $529 and stops counting.
