The activation funnel has a single chokepoint: step 3, connecting a data source. Of 5,000 signups, only 1,050 reached their first key action — a 79% drop. Users who clear step 3 retain at 31% at week 4. Users who skip it retain at 14%. That gap compounds: by week 12, completers are at 22% and skippers at 8%. Variant B closes this gap — +18pp activation rate, p=0.001. Ship it. Referral is your best acquisition channel: 31% week-12 retention versus 11% for paid search. Q2: fix the data connector UX and double down on referral.

Step 3 — connecting a data source — is where the funnel collapses. 550 users (34%) drop between onboarding complete and step 3. Another 215 abandon during the connection flow itself. “How do I connect my data?” accounts for 342 tickets (24% of total volume) with an 18-minute average resolution time. Users who clear step 3 don’t come back to support about it. Those who don’t become churn. The fix is not documentation. It’s reducing friction in the connector setup UI.

Referral users hit 31% week-12 retention — more than double paid search (11%) and 7pp above organic (24%). Every cohort in the last 6 months shows the same ordering: Referral > Organic > Paid Search. Referral users arrive with context — a peer already explained the value — and activate faster. Paid search users arrive with high intent but low understanding; they churn the moment onboarding feels hard. Every dollar on referral incentives buys a 2x better retained user than paid search.

Variant B: 50% activation. Variant A: 32%. That’s an 18pp lift on 2,400 users per arm, p=0.001, 95% CI 11–25pp. This isn’t a marginal improvement you deliberate over. It’s a step-function change.

At 1,600 new signups/month, +18pp means ~290 additional activated users every month. Over a quarter, that’s ~870 users who would otherwise churn silently at step 3.

The only open question is whether the lift holds at scale. The only way to answer that is to ship it and measure. Running another test while this one screams ship is burning users you’ll never get back.

Charts & Reports: 6.8 uses/week. Data Connectors: 0.4. That’s a 17x frequency gap. Users who find Charts & Reports stay. Users who never connect a data source churn. The usage distribution is your retention story in miniature.

Data Connectors is the lowest-used feature and the highest-support-volume category — 342 tickets last quarter. That’s not a coincidence. It’s a UX failure masquerading as a user education problem.

A help article won’t fix 0.4 uses/week. A redesign will. Reduce the connection flow from 7 steps to 3 and watch both usage and tickets move.

Three categories produce 52% of your tickets: data source connection (342), feature discovery (218), export & sharing (186). Fix one — data source connection — and you remove 24% of all support volume. Fix all three and you halve the ticket load.

Each category maps to a product gap, not a user gap. Data source connection → step 3 UX. Feature discovery → onboarding sequence. Export & sharing → workflow friction. Users aren’t failing to understand. The product is failing to guide.

These are sprint-sized fixes with support-cost-sized payoffs. Every week you don’t fix data source connection, you’re paying for 342 tickets that fix themselves when the UX does.

Sprint 1 — Ship variant B onboarding (p=0.001). Expected lift: +18pp activation rate on ~1,600 new signups/month — ~290 additional activated users per month.

Sprint 2 — Redesign the data connector UX. Target: reduce step 3 drop-off from 34% to under 20%. Proxy metric: “How do I connect my data?” support tickets. When tickets drop, the UX worked. This is the highest-impact product investment available.

Sprint 3 — Launch a structured referral program. Referral users retain at 2x. Growing referral from ~12% to ~25% of acquisition mix moves modelled week-12 retention from 24% to ≥28% without changing the product.

Ship variant B: 90-day retention moves from 24% toward 30%. +18pp activation means ~870 more activated users over 3 months. More activated users compound into more retained users.

Don't ship variant B: activation stays at 32%. Those ~870 users never activate. Week-12 retention stays flat at 24%. Same funnel, same results next quarter.

Fix the connector UX on top of variant B: reducing step 3 drop-off from 34% to 20% adds ~560 activated users per quarter. Combined with variant B, 90-day retention pushes toward 33%.

Grow referral to 25% of acquisition mix: this alone lifts overall week-12 retention from 24% to ≥28%. Combined with variant B and the connector fix, projected 90-day retention: 35–38%.

Do nothing: retention stays at 24%. Paid search keeps buying users who churn. The connector keeps generating 342 tickets a quarter. Nothing changes.