Who this is for
Store owners, growth leads and marketing teams who pay for traffic and watch it leave: the checkout loses people at the same step every month, the product page does not answer the questions customers ask, or a redesign changed conversion and nobody can say in which direction.
I worked on growth and experimentation at Getir from 2020 to 2021 and have run experiments on client stores since. What I bring is the engineering under the experiment: tracking that fires once, a test that renders without flicker, and the discipline to fix the slow page before testing the button colour.
Typical situations
- Add-to-cart is fine, checkout completion is not, and the payment provider says everything works.
- Traffic doubled after a campaign and orders did not; nobody knows whether the landing page or the offer is the problem.
- An A/B testing tool was installed, three tests ran, and the results contradicted each other.
- Mobile converts at a fraction of desktop and the team suspects the page, not the audience.
- Conversion tracking double-counts orders, so every result looks better than it is.
- The site is slow, and every optimisation idea is blocked by the question of whether speed is the real cause.
What I deliver
- Funnel analysis from real data: GA4, server logs, session recordings where consent allows, and the checkout’s own error logs, to find the step and the reason people leave.
- Fixing the infrastructure first: page speed, form and checkout errors, tracking that fires once, so an experiment measures the change and not the noise.
- Experiments with a written hypothesis, a sample size and a stop rule, implemented in the templates rather than in a client-side tool that flickers.
- Product page, cart and checkout improvements grounded in what customers ask: delivery and return information, stock, payment methods, trust and error messages.
- Landing pages for campaigns that match the ad, load fast and are measured against the same events as the store.
- A written test log the team keeps: what was tried, what it showed, what was kept.
How it goes
Measure before touching anything. I check that the events, the orders and the revenue in GA4 match the store’s own numbers, then map the funnel step by step. If tracking is wrong, that is the first fix.
Fix the obvious, then test the uncertain. Broken forms, slow pages and missing information are fixed without a test. What is genuinely uncertain becomes an experiment with a hypothesis and a stop rule.
Report and keep what works. Each experiment ends in a short written result. Winners go into the templates for good; losers are written down so they are not tried twice.