I Built a 2.8KB Cookieless Analytics Tool After Google Analytics Hid 35% of My Traffic

Azaan Ali Raza — Tue, 23 Jun 2026 19:57:49 +0000

Last month my Next.js SaaS showed 1,200 visitors in Google Analytics.

That's 35% of my traffic GA simply missed because of ad blockers and cookie consent rejections. I am 20, building from Meerut, and I was making product decisions on broken data.

So I deleted GA and built my own.

The problem
I launched AmazeMatrix in early 2026 with GA4 installed. Three things broke immediately:

The cookie banner killed UX. 60% of Indian users clicked "Reject". My bounce rate jumped.
Ad blockers ate my data. Over 40% of internet users run blockers that kill analytics cookies by default.
Speed. GA's script was 45KB. My LCP went from 1.9s to 3.1s on mobile.
I tried the privacy tools everyone recommends:

Plausible — $9/mo, super clean, but no heatmaps, no Core Web Vitals
Fathom — $15/mo, managed, but no AI insights, no India pricing
OpenPanel — $2.50/mo, great product analytics, but I still needed timeline annotations for deploys
I didn't want 4 tools. I wanted one lightweight snippet that gave me traffic, speed, and why things changed.

Investigation: how cookieless actually works
I spent a week reading how the good ones do it. Three patterns:

No persistent cookies. Instead of storing an ID for months, you generate a daily rotating hash from IP + user agent + salt that changes every 24h. You can count uniques today, but you can't track someone forever.
Server-side or edge capture. If the browser never sets a cookie, ad blockers have nothing to block.
Relative coordinates for heatmaps. Instead of sending screenshots (heavy), you send x/y as percentages of the element. That's under 0.2KB per click.

That was the blueprint.

Solution: I built AmazeMatrix
Tech stack I chose because I live in Next.js 16:

Frontend: Next.js App Router, Tailwind, Framer Motion
Edge: Cloudflare Workers / Vercel Edge
DB: Neon Postgres + Drizzle ORM
Auth: Better Auth
AI: Gemini SDK
Billing: Razorpay (because Stripe doesn't work for most Indian founders)
The entire snippet is 2.8KB gzipped:

3 lessons I learned building privacy analytics

Don't fingerprint, rotate.
I first tried a persistent hash. It felt creepy and broke privacy laws. Daily rotating hash gives me accurate daily uniques without storing PII. Privacy-first is not a marketing line, it's an architecture choice.
Heatmaps don't need images.
Everyone sends full-page screenshots. I send {x: 0.42, y: 0.18, el: '#pricing'}. That's 18 bytes. Multiply by 10k clicks and you save megabytes.
Founders don't want another dashboard, they want answers.
The most used feature is not the graph. It's the AI chat. "How is page speed affecting sales?" gets more clicks than any chart.

How you can try it
I kept it stupid simple:

I'm building in public on X and Threads as [@razavi_azaan]. If you want the self-host Docker compose, comment "self-host" and I'll DM it.

I thought 200 people visited my site. It was 11.

Azaan Ali Raza — Tue, 23 Jun 2026 11:00:59 +0000

3 AM. Launch night.

Google Analytics said 200 visitors. 12 signups. I was buzzing.

Then my co-founder messaged me:

"Bro your site isn't loading any scripts on my work laptop."

I checked. uBlock Origin — blocks GA. Privacy Badger — blocks GA. His company firewall — blocks GA.

I wasn't reading my traffic. I was reading the survivors.

I pulled the real number the next morning.

11 people.

Not 200. Eleven.

I had spent 6 hours that week A/B testing headline copy. Rearranging buttons. Tweaking CTA colors. All of it — based on data that was missing 95% of my actual visitors.

That's when I stopped using Google Analytics. And started building AmazeMatrix.

The idea was simple: analytics that can't be blocked.

No cookies. No third-party domain. Script loads from your domain as a first-party request. Blockers don't see it. Firewalls don't flag it. Under 2KB so it doesn't slow anything down.

First week of running it on a real site?

Actual traffic was less than half of what GA reported. Gone. Invisible. Making decisions off ghost data.

Building it wasn't clean either.

My "active visitors" counter hit 47 on a page with 8 real sessions. Bounce rate was firing before the page finished loading. Three beta testers churned silently — a CSP header was eating my script and nobody told me.

Every bug made me angrier at GA for normalising bad data.

I fixed all of it. Session expiry. SPA route tracking. A CSP handshake so you know the script fired before you trust a single number.

The tagline wrote itself:

Know more. Track less.

If you're building something and quietly suspect your analytics are lying to you — they probably are.

Early access is open → amazematrix.com

Drop your current analytics stack in the comments. Genuinely curious how many people have caught GA lying.

DEV Community: Azaan Ali Raza

I Built a 2.8KB Cookieless Analytics Tool After Google Analytics Hid 35% of My Traffic

I thought 200 people visited my site. It was 11.