Email Machine: an internal tool to a SaaS platform

Context

Email Machine began in 2009 as a custom sending tool for one of the region’s fastest-growing e-commerce companies. By 2013 it was clear the thing built for one giant (reliability, speed, data depth) was exactly what the rest of the market was missing. We co-founded the company in 2014; it grew to more than 100 million emails a month for 196 brands across five countries, and was acquired in 2024. I stayed on for product and UX.

Problem

In 2009 the email tools on the market could not take that scale or its integration demands: millions of transactional messages, deep catalog and order hooks, deliverability that held up during a holiday peak, and segmentation that did not need an engineer for every campaign. Taking it to market later, the problem shifted: smaller teams had the same needs and the same bad options. A generic tool that treated deliverability as their problem, or scripts with no way to test a subject line.

Approach

Build deliverability, segmentation, and A/B testing into the core, not as add-ons, and design from the campaign owner’s send day backwards to the infrastructure. Compete on the five things that matter rather than a feature list of fifty. Invest early in machine learning for segmentation, personalisation, and send-time optimisation: the shift from mass blasting to per-recipient timing was coming.

What I built

Scoped to my part as co-founder leading product and UX: the product direction, and the editor and automation builders that let a non-technical marketer run a complex campaign; the segmentation model over subscriber data and behaviour; the reporting dashboards; and the ML roadmap for personalisation and delivery-time optimisation. The sending path runs on Node.js, PostgreSQL and Redis with RabbitMQ for async work, on AWS with SendGrid/SES for delivery, and a custom event layer feeding Mixpanel.

Platform

Results

100M+ emails a month at peak, across CZ, SK, PL, HU and RO. 196 e-commerce clients including Astratex, DámeJídlo, Nespresso and Generali. Fifteen years in production; acquired in 2024. Measured by monthly send volume, client count, and the retention that followed the shift in what we optimised for.

What I’d tell you if you’re facing this

Three things. If deliverability, segmentation, and testing are the point of the product, they have to be in the first version: retrofitting them later means rebuilding the data model. Watch the metric you optimise: we moved from acquisition, a vanity number, to retention, and profitability followed. And the un-fun parts (bounce handling, spam-complaint loops, list hygiene) are most of the actual work; the customer conversations that surface them beat internal guesses every time.