About the job

Own the experimentation engine behind a $1Billion+ e-commerce brand's growth decisions.

Compensation: Up to AED 25,000/month (tax-free) + monthly performance bonus of up to 10% of baseLocation: Dubai, UAE (On-site) with relocation assistance and visa supportGrowth path from here is Senior Data Analyst → Analytics Engineer. What takes 4-5 years elsewhere happens in 1-2 here.

We've been AI-first since December 2023: Every analysis ships with an LLM validation link, and agentic workflows are in the analysis pipeline, not a pilot. You'd be building on a foundation, not evangelising one.

Role Overview:Puffy, a market-leading, nine-figure D2C brand and Top 5 US mattress brand with over $1Billion in sales, is hiring a Data Analyst (On-site in Dubai) to own the full experimentation program, from hypothesis to business recommendation, with ownership over test design, statistical analysis, data pipelines, and cross-functional delivery, reporting directly to the Analytics Manager.

This role exists to convert raw platform and site data into tested, statistically validated recommendations that marketing, UX, and development act on the same week, building the experimentation infrastructure and analytical rigor that let growth decisions run on evidence rather than intuition.

• The Problems You'll Own:Primary: Turn every experiment into a decision, not a readout: We run A/B tests across the e-commerce platform but the gap is in what happens after the data comes in. You take test results and turn them into clear, statistically solid recommendations: what worked, what didn't, why, and what to do next. You need to be able to go deep at the segment, device, and source level to find what top-line reporting misses.
• Secondary: Build the self-serve data infrastructure so experimentation scales: Our experiment data comes from first-party clickstream we built ourselves, millions of events a day joined against Shopify orders, ad platforms, and our own customer data in BigQuery. Bot traffic, identity stitching, and attribution gaps can quietly flip a result. You write the SQL and dbt models behind your own analyses, check the data before trusting it, and flag what's broken rather than reporting a number you can't stand behind.