Maven Fuzzy Factory: Marketing Funnel & A/B Tests
Marketing
Which channels convert, where does the funnel leak, and did the website A/B tests actually work?
Maven Fuzzy Factory is a simulated e-commerce startup selling plush toys online, launched in March 2012. Acting as its first data analyst, this project answers a year and a half of leadership’s real questions about marketing channels, bid optimisation, conversion funnel drop-off, website A/B tests, and seasonality, entirely in SQL.
Findings
Channel choice barely matters for new-customer conversion once repeat visitors are controlled for. gsearch and bsearch together account for about 72% of sessions, but restricted to first-time sessions, most channels converge to about 6.7-6.9% conversion.
Desktop converts far better than mobile, and billing is where visitors drop off. Desktop converts about 2.6x better than mobile on gsearch nonbrand (8.22% vs 3.18%, new visitors). Billing-to-order conversion is only 62.1%, meaning 38% of visitors who already entered shipping details abandon at payment.
The billing-page redesign is a confirmed win. The new /billing-2 page converts at 63.4% versus the original /billing’s 44.8%, an 18.6 point gain, about a 41% relative lift. Separately, /lander-3 was confirmed as a losing landing-page variant, converting at less than half the rate of /lander-2.
Live dashboard
Full write-up
- Full report — methodology, limitations, and recommendations
- SQL and notebook on GitHub