How we instrumented our own SaaS with CrazyLeads
We installed our own pixel inside our own app: logged-in user identify, device merging by email, and product events on the same timeline as marketing. What worked, what broke, and what we learned about tracking a SaaS.
Every data-tool vendor says their product works for SaaS. We decided to prove it the most honest way there is: installing our own pixel inside our own app — the entire logged-in area, from sign-up to daily use. This is what we built, what broke along the way, and what stuck as lessons for anyone who wants to understand how the users of a SaaS behave.
What a SaaS needs that a website does not
Tracking a sales site means measuring traffic that converges on one event: the purchase. Tracking a SaaS is a different question — who signed up, who activated, who uses what, who went quiet. Acquisition data (which ad brought the person) and product data (what they did after logging in) must live on the SAME person, or the funnel breaks exactly at the turn that matters most: anonymous → account.
The three pieces we installed
- 01The pixel across the whole app — public pages and the logged-in area. Single-page React application: the pixel emits a pageView per route on its own, no configuration.
- 02Identify on the logged-in session: every authenticated navigation tells the pixel the account email. All previous anonymous browsing — including the visit that came from an ad — merges into the person.
- 03Product events on the milestones that matter: sign-up completed, project created, source connected, delivery created, report published. One line of code per event.
Results showed up in minutes, with a proof we did not expect so soon: the second identified user came in through two different devices — and the platform merged both into the same person by email. Their timeline shows the whole journey: the pages on the laptop, the login on the desktop, the product events. That is the picture no anonymous-session tool can give you.
What broke (and why it is worth telling)
Using your own product exposes what no demo ever will. In one day of dogfooding we found and fixed three real defects — one made new projects be born without the identity graph, another kept an identified email out of the leads list. None of them showed up for customers on the classic paid-media path; all of them showed up on day one of the SaaS path. A tool the company itself does not use on a given use case is a tool untested on it.
The most transferable lesson: in a logged-in app, TURN OFF automatic form capture. The "invite a teammate" form carries someone else’s email — auto-capture would overwrite the session identity with it. In SaaS, identity comes in through identify, not through forms.
What we can answer now
- Which channel brings the sign-ups that ACTIVATE — not just the ones that fill the form.
- How long passes between the first ad seen and the first project created.
- Which features the users who stay touch in week one — and which ones the users who vanish never touched.
- All of it via SQL and publishable artifacts — or by asking Claude in plain language, connected via MCP.
Where we are honest about the limit
For retention funnels with weekly cohorts and path analysis, a dedicated product-analytics tool still ships more out of the box — we ship the unified data, the identity and the SQL to build the analysis. For many SaaS companies that is the right trade: one instrumentation serving product, marketing and activation, instead of three tools with three truths. For the rest, CrazyLeads plays along: PostHog, Mixpanel and Amplitude are among the Delivery destinations.
The identify and product-event walkthrough lives in the pixel installation guide in our docs. The app you use at CrazyLeads is, right now, measuring itself with everything described here.