your checkout page is a leaky bucket. everyone knows the bucket leaks. the useful number is where.
i'm om yaduvanshi, 18, an applied ai developer building cloudline, a cookieless analytics tool. day 18 of 90, building in public.
the funnel is the whole product
a funnel is the most honest view in analytics. it takes pageviews and clicks, lines them up as steps, and shows you where people drop off. landing, pricing, checkout. the page where the line falls off a cliff is the page you fix first.
cloudline builds funnels out of pageviews and button clicks. you pick the steps, it shows the drop-off between them. built in, nothing to configure.
what funnels usually need
most analytics tools do funnels with identity. they stitch sessions together so they can tell you the person who hit your pricing page tuesday came back thursday and bought. that stitching is exactly what cookies do, and it's exactly why the consent banner exists, and the banner eats your numbers. every visitor who declines vanishes.
cloudline stores no IPs, sets no cookies, doesn't read user agents. a pageview is four things: page, referrer, screen size, time. so there is no stitching. i can't tell you if tuesday's visitor came back thursday. said that on day 17 too. still true.
the page-level funnel
so what can four data points do? pages arrive in order. a funnel of pageviews asks a simpler question: of the people who saw step one, how many saw step two. the same person visiting twice counts twice. i'm not hiding that. it's the trade, and it's written down.
button clicks work the same way. clicks are stored per page with the label you chose, so "saw pricing" to "clicked start trial" is a funnel with no identity in it at all. the drop-off between those two steps is the most actionable number in the whole dashboard. nobody argues about what it means.
pair it with scroll depth
funnels tell you where. scroll depth tells you why, sometimes. 25, 50, 75, 100 percent, per page. if 80 percent of visitors never get past 25 percent of your pricing page, your funnel drop-off isn't a mystery. your headline is.
neither of these knows who anyone is. that's the point of the whole tool. the page where people give up is a page problem, not a person problem.
free while in early access. no card.
— om
Top comments (2)
With no stitching, I'd label this as a ratio of page/event counts rather than "of the people who saw step one, how many saw step two." A small fixture makes the difference visible: 10 pricing views, 8 direct checkout views from an email, and 3 repeated trial-button clicks. None of those counts establishes that a pricing visitor progressed. Does the UI show the raw counts and counting window, and allow the next-step ratio to exceed 100% without calling it a conversion rate? That would preserve the privacy trade without implying a journey the data can't identify.
this is the sharpest framing of the no-stitching problem i've seen.
you're right: a funnel here is a ratio of event counts, not people. 10 pricing views and 8 checkout clicks from an email: 8/10. on the days it lands at 12/10 because three people hammered the trial button, it stays 120%. i'd never cap it at 100 or call it a conversion rate. the weird days stay weird, that's the honest version.
thanks for the fixture by the way. might borrow the 10/8/3 example.