All case studies / Websites & Landing Pages

Two A/B experiments on the lead form: an honest report with no winner, original kept on the site

Industry: Children's sports schoolTimeline: June 26 – July 11, measured for 7 daysArea: Websites and landing pages

2 experimentsquiz form and offer near the button, traffic 50/50

3,742 visitsduring the week-long measurement

p = 0.23 and 0.43difference within statistical noise, no winner

I need something similar

Before

Goal: increase conversion to applications for a free class
Agreed on two hypotheses to test as a proper A/B experiment
H1, form fields, spam protection, and the native form submission could not be changed

What we built

Three-step quiz: age, city and branch, contact details; the child's age is sent with the application
Offer block near the first-screen button
B variants as standalone JS snippets via Yandex Experiments, variant A stays unchanged
Measured using the Yandex Metrica goal "Application for a free class", traffic 50/50 for a week
The experiments were stopped after the measurement

How it works

Visitorsplit 50/50

→

Variant BJS snippet via Yandex Experiments

→

Lead formnative submission

→

Metrica goalapplication for a class

→

Reportconversion and p-value

Result

Final report sent, completion certificate issued
Experiments stopped, original version left on the site
After the report, the client ordered the next cycle

Technology

JavaScriptYandex MetricaYandex Experiments (Varioqub)MODXPlaywright

What matters in a project like this. During the first few days, variant B was not shown to anyone: the experiment flag comes as an array inside the flags field, while the loader compared it to a string. An end-to-end browser check caught this the same day, so the measurement window was not affected.