The Poor Man's Analytics Stack: How I Set Up Event Tracking in 2 Days With Zero Budget

The problem hit me on a Tuesday afternoon

We were three weeks into building Sonic Linker's core product. Our AI linking platform was starting to work, we had our first pilot lined up for Friday, and I suddenly realized we had no idea how anyone was actually using the product.

I needed to track: - Which AI suggestions users were accepting vs. rejecting - Where people dropped off in our onboarding flow - How long it took to create their first project - Whether our "smart recommendations" were actually smart

The standard answer? Hire a data engineer, set up Segment, pipe everything to a warehouse, build dashboards in Looker. Cool. That would take three months and $50k we didn't have.

I had 48 hours.

What I actually built (and why it worked)

Here's the stack I put together:

PostHog (self-hosted community edition) for event tracking. Free, open source, and you can spin it up on a $20/month VPS. I deployed it on Railway in about 30 minutes. The UI isn't as pretty as Amplitude, but it captures events, builds funnels, and lets you replay sessions.

A single events.js file in our frontend. 50 lines of code. Every important action called one function: `trackEvent('action_name', {properties})`. That's it. No complex taxonomy, no event schema debates, just ship it.

Google Sheets as my data warehouse. Laugh if you want, but I set up a simple script that pulled PostHog's API data every morning at 6am and dumped it into Sheets. My co-founders could see the numbers without learning SQL. Took me 20 minutes to write.

Metabase for the one fancy dashboard we needed for investor updates. Connected directly to PostHog's Postgres database (which is just sitting there, ready to query). Free tier, looks professional enough, took two hours to set up basic charts.

Total cost: $20/month for hosting. Total setup time: about 6 hours across two days.

The three rules that made this actually useful

I've seen teams spend months building "proper" analytics that nobody uses. Here's what I learned at Sonic Linker that kept our scrappy setup valuable:

Track outcomes, not actions. I didn't track "user clicked button." I tracked "user created first AI suggestion" and "user accepted AI recommendation." The difference matters. One tells you what happened, the other tells you if you're solving the problem.

When we realized only 23% of users were accepting our AI suggestions in the first session, that was a product problem, not a tracking problem. We fixed the product, not the dashboard.

Weekly review, not real-time panic. I picked five metrics that mattered (activation rate, time to first value, suggestion acceptance rate, weekly retention, and drop-off points). Every Monday morning, I looked at these for 15 minutes. That's it.

No Slack alerts, no real-time dashboards on TVs, no 3am freakouts about a metric dip. Just consistent, weekly pattern recognition.

Make it accessible to non-technical people. My co-founder (who codes) could dig into PostHog. My other co-founder (who doesn't) got a Google Sheet every Monday with the five numbers and a two-sentence summary of what changed.

Fancy tools are worthless if only one person can use them.

What I'd do differently now (but only slightly)

If I were starting today, I'd probably use Plausible or PostHog Cloud instead of self-hosting. The $20/month I saved wasn't worth the one Saturday I spent debugging a database connection issue.

I'd also set up basic alerts earlier. Not for every metric, but for critical drops. When our activation rate dropped from 45% to 28% over three days, I only noticed on Monday. By then, we'd onboarded 30 users into a broken flow.

But honestly? The scrappy stack worked. Three months in, we had clear data on what users actually did, where they struggled, and what features mattered. We made product decisions based on real behavior, not gut feel or whoever talked loudest in standup.

The real lesson isn't about tools

At Finvestfx, we had a "proper" setup with Mixpanel, a data team, and dashboards for everything. We still made bad decisions because we tracked everything and understood nothing.

At Sonic Linker, we tracked five things that mattered and checked them weekly. We made better decisions with worse tools.

You don't need a data engineer to set up analytics. You need clarity on what you're trying to learn and the discipline to look at the same few metrics consistently.

The rest is just plumbing. And plumbing is surprisingly easy when you stop trying to make it perfect.

Start here: Pick the three questions you'd ask users if you could talk to all of them today. Those are your metrics. Find the cheapest tool that tracks them (PostHog, Mixpair, even just logging to a database and counting rows). Set a weekly 15-minute calendar reminder to review them.

You'll have working analytics by Friday.