quizVariant identify trait (like-for-like)stxt-490006 · dbt marts d2c_prod_dbt_stxt_analyticsEach row scales the longer bar to full width, so the gap you see is the gap in the data. The chip on the right is the multiple (e.g. “2.6× higher” = creator_funnel_v2’s rate is 2.6 times match_me’s). Rates are % of each funnel’s signups (match_me n=47, creator_funnel_v2 n=54).
landerHomeV2) isn’t firing (1 recorded visitor vs 54 signups — GTM-1988). Since all paid traffic routes through Everflow, we use Everflow clicks per Offer URL as the visitor proxy: match_me (Main Page #413) = 3,202 clicks → 47 signups ≈ 1.5%; creator_funnel_v2 (quiz_v2 #422) = 3,377 clicks → 54 signups ≈ 1.6%. Near-identical click volume and click→signup rate (~1.1×); creator_funnel_v2’s advantage emerges after signup (the steps above). Caveat: Everflow clicks are paid traffic only; for reference, the on-site match_me lander (landerHome) shows 1,538 page views → 47 (3.1%).
Two acquisition funnels run on stxtai.com: a visitor goes through a quiz and is sent to justsext.com to create an account. match_me is the original path; creator_funnel_v2 is the new /quiz_v2 path. Both are new and were tested concurrently.
Cohort: users who created an account May 28–Jun 3, keyed on user_id, with the funnel taken from the quizVariant identify trait (dim_user_quiz_preferences) — the one signal recorded identically for both funnels. Behavior is read from fct_user_journey_events.
The app emits events through Segment (analytics.js). Segment writes to BigQuery (d2c_prod); dbt builds the marts behind the team dashboards. The same stream also feeds Mixpanel.
/quiz_v2=18, /quiz/default=11, unknown=3); cohort sizes match the day-level signup mart (47 / 54–55) and Mixpanel (42 / 53) within ~5–10%; paid = 0 for both in Mixpanel and the warehouse.
| Metric | match_me | creator_funnel_v2 | Gap |
|---|---|---|---|
| Signups in window (count) | 47 | 54 | 1.15× |
| Click → signup (Everflow paid-click proxy) | ~1.5% (3,202→47) | ~1.6% (3,377→54) | ≈1.1× |
| Started a chat | 47% (22) | 57% (31) | 1.2× |
| Received in-chat photo | 28% (13) | 61% (33) | 2.2× |
| Chat-start events / signup | 2.5 | 3.5 | 1.4× |
What this means: the two funnels brought a similar number of signups (47 vs 54 — driven by traffic split, not conversion). Once inside, creator_funnel_v2 users engage more: a higher share start a chat, more than twice the share receive an in-chat photo from the AI, and they generate more chat activity per user. (Signup counts are volume, not a conversion rate; chat-start events repeat on reconnect; in-chat photo is a server delivery signal.)
| Step | match_me | creator_funnel_v2 | Gap |
|---|---|---|---|
| Hit demo paywall | 9% (4) | 15% (8) | 1.7× |
| Clicked “Subscribe” | 6% (3) | 17% (9) | 2.6× |
| Reached checkout | 2% (1) | 7% (4) | 3.5× |
| Paid subscription | 0 | 0 | — |
| Token purchase | 0 | 0 | — |
| Returned on a later day | 2% (1) | 6% (3) | 2.6× |
What this means: creator_funnel_v2 users move further down the paid funnel at every step (paywall → Subscribe click → checkout), and the relative gap widens deeper in the funnel (1.7× → 2.6× → 3.5×). Neither funnel has produced a paid conversion yet; in this data paid lands a median ~35 days after signup, and this cohort is ≤7 days old. Return-visits favor creator_funnel_v2 (6% vs 2%) but on tiny counts (3 vs 1).
1. Small samples. 47 and 54 signups; paywall/checkout/retention are single-digit counts. Read as directional, not statistically conclusive.
2. Both arms use the identify trait. The funnels are instrumented differently — match_me stamps its variant onto events; creator_funnel_v2 writes it only to the user profile. To compare like-for-like, both are keyed on the quizVariant identify trait (the only signal recorded the same way for both). Tracked in GTM-1985.
3. Visitor→signup uses an Everflow click proxy. On-site page views can’t compare the funnels because the V2 lander page view (landerHomeV2) isn’t firing (1 visitor vs 54 signups — GTM-1988, fix in progress). Proxy via Everflow clicks per Offer URL (paid traffic): match_me (#413) 3,202 clicks → 47 signups ≈ 1.5%; creator_funnel_v2 (#422) 3,377 → 54 ≈ 1.6% — near-identical. For reference, the on-site match_me lander (landerHome) shows 1,538 page views → 47 (3.1%); Everflow conversions per Offer URL are also in BigQuery (everflow.conversion_report).
4. Purchases need more time. Median signup→paid is ~35 days; a ≤7-day cohort can’t show paid yet. 0 paid is expected at this age.
5. Retention is right-censored. Most signups landed in the final 2 days of the window, and the journey-event log excludes some in-app actions (e.g. message_sent, image_generated). Treat retention as early/partial.
-- Cohort + behavior, both arms keyed on the quizVariant identify trait
WITH acct AS (
SELECT user_id, MIN(event_date) AS signup_date
FROM d2c_prod_dbt_stxt_analytics.fct_user_journey_events
WHERE event_name='account_created' AND event_date BETWEEN '2026-05-28' AND '2026-06-03' AND user_id IS NOT NULL
GROUP BY user_id
),
coh AS (
SELECT a.user_id, a.signup_date, d.quiz_variant
FROM acct a JOIN d2c_prod_dbt_stxt_analytics.dim_user_quiz_preferences d USING (user_id)
WHERE d.quiz_variant IN ('match_me','creator_funnel_v2')
),
ev AS (
SELECT coh.quiz_variant, coh.user_id, coh.signup_date, e.event_name, e.event_date
FROM coh JOIN d2c_prod_dbt_stxt_analytics.fct_user_journey_events e ON e.user_id = coh.user_id
)
SELECT quiz_variant,
COUNT(DISTINCT user_id) AS cohort,
COUNT(DISTINCT IF(event_name='chat_initiated', user_id, NULL)) AS chat,
COUNTIF(event_name='chat_initiated') AS chat_events,
COUNT(DISTINCT IF(event_name='escalation_photo_sent', user_id, NULL)) AS photo,
COUNT(DISTINCT IF(event_name='onboarding_paywall_hit', user_id, NULL)) AS paywall_hit,
COUNT(DISTINCT IF(event_name='onboarding_paywall_click_subscribe', user_id, NULL)) AS subscribe_click,
COUNT(DISTINCT IF(event_name='checkout_form_rendered', user_id, NULL)) AS checkout,
COUNT(DISTINCT IF(event_name='subscription_paid', user_id, NULL)) AS paid,
COUNT(DISTINCT IF(event_date>signup_date, user_id, NULL)) AS returned_later
FROM ev GROUP BY quiz_variant ORDER BY quiz_variant;
-- Visitor -> signup by funnel, from lander page views (page_name + quiz_variant)
SELECT page_name, quiz_variant, COUNT(DISTINCT anonymous_id) AS visitors
FROM d2c_prod.page_view
WHERE DATE(timestamp) BETWEEN '2026-05-28' AND '2026-06-03'
GROUP BY 1,2 ORDER BY visitors DESC;
-- landerHome / match_me = 1538 visitors -> 47 signups (~3.1%)
-- landerHomeV2 / creator_funnel_v2 = 1 visitor (page view not firing) despite 54 signups -> unmeasurable (GTM-1988)
-- Visitor proxy via Everflow clicks (paid traffic) — covers both funnels
-- Everflow Reporting API: POST https://api.eflow.team/v1/networks/reporting/entity
-- {"from":"2026-05-28","to":"2026-06-04","columns":[{"column":"offer_url"}],"query":{"filters":[]}}
-- total_click by Offer URL: #413 Main Page (match_me) = 3,202 -> 47 signups (~1.5%)
-- #422 quiz_v2 (creator_funnel_v2) = 3,377 -> 54 signups (~1.6%)
Prepared by GTM analytics · single source: BigQuery stxt-490006 via dbt marts. Renders in light & dark. To refresh: re-run the appendix SQL with a new date window and update the figures + the “Data as of” badge.