SYS/CASE
Growth & Performance · CRO · SEA · Analytics · 2022–2023
Institut Cassiopée — +120% conversion in three engagements: UX, SEA, dashboards
By Mathieu Milon, AI engineer growth & CRO · Published
140 pages, tens of thousands of visits a month. Three engagements — UX, SEA mini-site, dashboards — for the same client, in tabs rather than three siloed pages.
+120% conversion
Context
Institut Cassiopée is a continuing-education body (cassiopee-formation.com). A 140-page site, tens of thousands of visits a month, and a clear commercial goal: turn traffic into training-quote requests.
Problem
Visitors explored content without ever entering the training → quote tunnel. Paid traffic converted even worse than organic: same audience, but not the same need or the same landing page.
Approach
Three engagements over two years, shown here as tabs: UX/CRO optimisation of the site, a mini-site dedicated to paid traffic, and the analytics instrumentation that made it possible to steer the tests.
UX / CRO
After tracking, UX/SEO recommendations to turn the visitor into a prospect.
Test protocol in detail
Heuristic grid (friction, clarity, value, motivation), ~15 hypotheses prioritised with PXL, 8 months of A/B tests on Google Optimize, stop at 99% significance.
+120%
overall CVR
60%
of tests with lift ≥ 10%
+30%
engagement
PXL prioritisation of test hypotheses Optimised quote-request page
SEA mini-site
Paid traffic (Google / Meta) explored without entering the tunnel: people were looking for a course, not the catalogue.
Mini-site in detail
Response: 9 dedicated landing pages on Webflow, templates tested in A/B then MVT, YouTube videos loaded on click so they would not tank load time.
+65%
CVR vs site, paid segment
+22%
engagement
Paid mini-site landing page
Dashboards
Tagging plan, event listeners (CTA, account, menus, scroll, video), Lucky Orange and Looker Studio dashboards — Meta Ads connectors included.
Instrumentation in detail
Universal Analytics → GA4 migration to keep history readable.
Conversion funnel in Looker Studio KPI dashboard, Meta Ads connectors included
Where AI comes in
2022–2023 engagement: analysis and prioritisation by hand (PXL). Today the same tagging plan is generated then reviewed, and extract analysis is done in SQL + Python in days rather than weeks.
Results
+120%
overall CVR (UX/CRO)
60%
of tests ≥ +10%
+30%
engagement
+65%
paid CVR (mini-site)
| Measure | Before | After |
|---|---|---|
| UX/CRO conversion | 100 | 220 |
| Paid conversion (SEA mini-site) | 100 | 165 |
| Engagement | 100 | 130 |
Base 100 = measure before intervention · +120% conversion (8 months, ~15 PXL hypotheses, 60% of tests ≥ +10%) · +65% on paid, 9 landing pages · +30% engagement.
Takeaway
- 1
We stop a test at 99% significance, not at the first encouraging signal: statistical discipline protects against false positives.
- 2
Paid ≠ organic: it is not the same tunnel. A page dedicated to paid traffic converts better than a site designed for organic.
- 3
The dashboard is there to steer tests, not to decorate: without clean instrumentation, we optimise blind.
A similar project?
Let's talk about your next engagement
A funnel, a product or an extract to analyse: book a 30-minute call, written scoping within 48h.