i'm om yaduvanshi. i'm 18, an applied ai developer building cloudline, a cookieless google analytics alternative. one script tag, under 2 KB, no cookies, no stored ips.
funnels are the one analytics feature that seems to need identity. a funnel says: 100 people saw the pricing page, 40 clicked start free, 12 signed up. to connect those steps you usually follow one person through the whole journey.
here's the problem. cloudline can't follow anyone.
strip identity out and a funnel stops describing individual visitors. it just describes the counts.
cloudline counts pageviews and button clicks per page. a pageview stores four things: page, referrer, screen size, time. that's it. no cookies, no stored ips.
a funnel is those counts stacked in the order you care about. pricing page got 100 views. the start free button got 40 clicks. the signup page got 12 views. the drop between steps is the same drop. it was never attached to a name.
without identity, the funnel can't guess your steps. a cookie-based tool watches a person move from page to page and draws the path for you. cloudline can't do that, so you pick the steps: this pageview, then this button click, then that pageview.
i decided that was fine. the automatic version mostly invents paths you never designed anyway.
button clicks fill in the middle of the funnel. add data-cloudline-track to the button and only the label is stored. clicks on "start free" sit between the pricing pageview and the signup pageview, which pageviews alone would never show.
a funnel with no identity has one honest caveat. it assumes the 40 people who clicked are inside the 100 who viewed. they probably are. sometimes they aren't. a visitor can click start free, leave, and come back tomorrow. cloudline counts that as one click and two visits. no way around it.
i show the counts anyway because the direction of the drop matters more than the exact math. if 100 become 40 become 12, the shape of the problem is obvious. you don't need a person-level trail to know the pricing page leaks.
i look at one funnel per important path. not ten. one.
if the drop is between step one and two, the page doesn't sell the click. if it's between two and three, the button works and the next page doesn't. the funnel points at the problem. it doesn't diagnose it. live visitors help here. when i change a page i watch today's visits in real time and see whether the counts start moving. same anonymous counts, just faster.
you lose the per-visitor journey. no "user 482 bounced at step 2". what you get is the shape of the funnel with zero tracking baggage: no cookies, no stored ips.
for an indie tool, that trade was easy. i'd rather explain one honest caveat than maintain an identity graph. i wrote about the other cookieless stitching problem last time: https://dev.to/omyvnss/how-i-track-scroll-depth-without-cookies-jb9
cloudline is free for early users while i keep building it. cookieless analytics in one line of code.
— om yaduvanshi