We Built a Decision-Making Tool: 30 Days of Funnel Data

For the last 30 days we've been building TangoEra — a decision-support tool that turns a short structured assessment into a visual "decision profile": how you weigh risk, time horizons, compensation vs. growth, and the other dimensions that actually drive your big calls.

This is the honest build-in-public report: what we shipped, what the funnel data says, and the uncomfortable lesson week one taught us.

What we built first

We went against the usual indie playbook. Instead of shipping fast and instrumenting later, we spent the early weeks on the unglamorous layer:

• A 7-dimension assessment (quiz) with per-step telemetry, so we can see exactly where people drop off
• A results/paywall flow with UTM-tagged sources (blog CTA, embedded quiz, result page) so every channel is separately measurable
• Three lead-capture pages wired into the same event pipeline
• An analytics-events worker streaming into a queryable dataset (we learned the hard way that eventually-consistent key-value lists lie to you at small volume — direct SQL on the event table is the only read we trust now)
• An SEO/content layer: a decision-framework blog, Open Graph cards, hreflang for six languages

By launch-plus-one-week the engineering checklist was 79/79 green, 20 Pinterest pins were scheduled, and a 232-URL audit of the site came back clean.

Then we opened the window and measured

We gave the funnel a clean measurement window (T0 + ~28 hours of stable, fully instrumented production) and pulled the authoritative dataset directly.

External funnel events: 0.

Not near-zero. Zero. The daily event ledger confirmed it: every single event in the window traced back to our own smoke tests and probes. No quiz completions, no CTA clicks, no leads — from real visitors — in the entire window. Site UVs sat in the "hundreds per day" range on a good day, and the distribution channels we'd stood up (Pinterest, directories) hadn't produced their first outbound click yet.

The lesson: we built a product and forgot a distribution engine

Here's the part that stung. The conversion copy was fine. The instrumentation was fine. The drop-off analysis we wanted to run? Statistically impossible — you can't analyze a funnel with n = 0.

When the sample size is zero, every A/B test, every headline tweak, every paywall experiment is theater. We had optimized a machine that nothing was flowing through.

So week two got re-planned around a single sentence from our review doc: the funnel experiment is short on traffic, not copy.

What changed:

• Distribution became a first-class workstream, not an afterthought — one owner, daily cadence, its own backlog
• Content moved from "support docs for SEO" to an actual channel: decision-framework articles targeting long-tail queries, each with a hard CTA into the quiz
• We started measuring channels individually (separate UTM sources per surface) so that when traffic does arrive, we already know which door it came through
• We set explicit kill criteria: a channel that can't produce outbound clicks in 30 days gets cut, no sentiment involved

Where the data actually helped

Zero is a useful number. Because the instrumentation was solid before traffic existed, we can state the null result with confidence instead of wondering if the analytics are lying. We also caught a real bug class early — our first read of the "zero" was itself wrong (a stale list API masked 11 probe events that should have been there), which taught us to verify reads with strong-consistency queries before making decisions on them.

That meta-lesson — instrument the instrument — turned out to be worth the week by itself.

What's next

• Ship the distribution backlog (channels already live: Pinterest, startup directories, guest posts on decision-making content)
• Re-open the funnel window with the same rigor once inbound traffic crosses a threshold where numbers mean something
• Publish the week-2 review the same way: real data, no varnish

If you're building something too and you've also stared at a dashboard full of zeros — you're not behind, you're just at the step most people skip over. The fix isn't more features. It's more doors.

---

Follow along — we're posting these reviews weekly. And if decision-making content is your thing, the TangoEra quiz is free: three questions, one visual decision profile.