I Set Up Product Analytics With No Engineering Help. Here's What Actually Worked.

When I joined Sonic Linker's founding team, we had one engineer and about 47 things that needed building. Product analytics was priority number 23.

But here's the thing: you can't build an AI SaaS product blind. I needed to know if people were actually using the link generation feature. I needed to see where they dropped off. I needed conversion data before our next investor update.

The engineer had exactly zero hours to spend setting up a data warehouse. So I did it myself.

Start With Events You Can Actually Action

I see PMs make this mistake all the time. They instrument everything. Every click, every hover, every time someone breathes near the product.

I started with five events: - Account created - First link generated (our core value prop) - Link clicked (did it actually work?) - User returned within 7 days - Upgrade clicked

That's it. I used Mixpanel's free tier, dropped their JavaScript snippet in the header, and wrote the event calls myself. Took about three hours because I had to relearn JavaScript syntax, but it worked.

At Finvestfx later, I did the same thing for our forex treasury platform. Except there I cared about different events: report generated, API key created, first transaction logged. The pattern was the same though. Pick the 5-7 events that map directly to your product's core loop.

If you can't action the data within a week, don't track it yet.

Your CRM Is Half Your Analytics Stack

This sounds obvious but I missed it for two months at Sonic Linker.

We were using HubSpot for our early enterprise outreach. I was using Mixpanel for product events. And I was manually cross-referencing them in Google Sheets like some kind of data masochist.

Then I realized HubSpot has a workflows feature. I could trigger CRM updates based on product events. When someone generated their first link in Mixpanel, HubSpot would tag them as "activated." When they hit 10 links, they'd get auto-enrolled in an upgrade nurture sequence.

Suddenly I had behavioral segmentation without writing a single SQL query.

At Finvestfx, we used Zoho CRM. Same concept. I set up webhooks so that when a client generated their first compliance report (tracked in Mixpanel), their account owner got a Slack ping and a CRM task to check in. Our retention went up 18% in two months, not because the product changed, but because we were talking to users at the right time.

Your CRM already tracks accounts, deal stages, and revenue. Your product analytics tool tracks behavior. Connect them with Zapier or webhooks and you've got 80% of what a data warehouse would give you.

Google Sheets Is Underrated for Early Cohort Analysis

I know this sounds janky. It is janky. But it works.

Mixpanel's free tier didn't have great cohort retention charts. The paid tier was $800/month and we were a pre-seed startup eating instant noodles.

So I exported weekly signup cohorts to Sheets. Added a column for "active in week 1," "active in week 2," etc. Wrote some basic formulas to calculate retention percentages.

It took me 30 minutes every Monday morning. But I had clean retention data to share in our team meetings. I could see that our Week 1 retention was 34% (terrible), but Week 4 retention for activated users was 71% (pretty good).

That insight changed our onboarding flow. We stopped optimizing for signups and started optimizing for that first link generation. Activation rate went from 42% to 61% in five weeks.

You don't need Amplitude's fancy retention charts to see the pattern. You need discipline to look at the data every week.

The Stack That Actually Worked

Here's what I was running at Sonic Linker with zero data engineering: - Mixpanel (free tier, self-instrumented events) - HubSpot (CRM with workflow automation) - Google Sheets (weekly cohort exports, manual but fast) - Hotjar (session recordings for the top 3 drop-off points) - Zapier (glue between everything)

Total cost: $0 for the first three months, then about $150/month once we hit Mixpanel's event limits.

At Finvestfx, I added Metabase on top of our Postgres database once we had more budget. But honestly, the Mixpanel plus CRM combo handled 90% of decisions for the first year.

What I'd Tell My Past Self

Don't wait for the perfect data stack. You'll be waiting until Series B.

Pick 5-7 events that matter. Instrument them yourself. Connect your product analytics to your CRM. Export to Sheets when you need custom analysis.

I've seen PMs wait months for engineering to build a data pipeline. Meanwhile they're making product decisions based on vibes and the two users who email them.

You can set this up in a weekend. It won't be beautiful. But it'll tell you if people are actually using your product, and that's worth more than a perfect Snowflake schema.