Prom.ua: Leading a 100M+ Marketplace in the Age of AI (Expert Voices & Tips)

On this episode of the Retention Podcast, we sat down with Mykyta Artemchuk, Chief Product Officer at Prom.ua (EVO Group) — easily one of our most engaging guests to date. He’s quick with a joke, but behind the humor is the guy running product at one of Ukraine’s biggest online marketplaces.

Prom.ua is no small operation. We’re talking over 100 million products from 700,000+ sellers, attracting 30–50 million monthly visitors. It’s a heavyweight in Ukrainian e-commerce, and Mykyta shared invaluable insights about building products at this scale.

Here are the highlights from our chat.

Two Playbooks for Running a Product Office

Globally, product management tends to fall into two camps.

In the first model, product managers focus almost entirely on understanding users, shaping ideas into testable hypotheses, and running experiments. Then, they hand off requirements to dev teams they don’t directly manage.

In the second model — used by Google and Amazon — product managers command dedicated resources: their own teams, infrastructure, and direct influence over priorities. In this setup, PMs are essentially administrative leads for product squads of typically 7–15 people, including engineers, designers, analysts, and more. These squads are grouped by function (think search, mobile apps, company sites) with senior product leads overseeing clusters of them.

Neither model is a universal best — it’s more about finding the right fit for how your teams work.

The Three Metrics That Really Matter

Prom measures success through three key metrics. 

First is gross merchandise value (GMV) — the total value of goods and services sold on the platform — and its year-over-year growth. That’s the north star. The other two are revenue growth and profitability. Everything the product org does ties back to improving one of these numbers.

Big pushes, called “missions,” are planned quarterly and kept focused, usually five or six at maximum. Think launching free delivery subscriptions (a la Amazon Prime), adding pick-up points through another marketplace partnership, or rolling out buy-now-pay-later options. These missions often pull in multiple teams at once, and the structure is designed to make that cross-team collaboration less painful and more productive.

From Browsers to Apps — and the Next Leap

E-commerce’s evolution has been pretty clear-cut: first, desktop browsers and bookmarked sites. Then mobile, but with basically the same buying flow. Then came apps, and that’s where things changed.

Once shoppers install a marketplace app they trust, they’re far less likely to buy elsewhere. Without your own app, reaching them becomes either near impossible or ridiculously expensive. For Prom, building and promoting their mobile app transformed from “nice to have” to a survival move, sparking a complete rethink of the entire product experience and brand presence.

AI Search Is Shaking Things Up

Traffic from AI-powered search tools like ChatGPT is already showing up in the numbers and rewriting the rules. Just like SEO rewired the internet years ago, there’s now a race to figure out how to make products and platforms show up inside LLM results.

Obviously, it’s not about keyword stuffing anymore. Success requires managing brand reputation, structuring content for AI comprehension, and ensuring these models recognize you as a trusted source. Google’s shift toward AI-driven search and others experimenting with monetized recommendations mean the playbook will keep evolving fast.

AI as the Next Big Interface

Today, AI isn’t the all-knowing oracle some people imagine. It’s more like a new interface to the internet — with some rough edges but huge potential.

Case in point: OpenAI’s $6.5 billion acquisition of Jony Ive’s design company, before it had even released a product. The race isn’t just about building better models; it’s about owning the way people interact with them.

Recommendations: Helpful, Not Magical

In e-commerce, AI recommendations can move the needle, but they’re not magic. They need two things for improved performance: the right tech and vast amounts of quality data. Even with massive product catalogs, some categories just lack traffic for effective AI learning.

Prom’s own recommendation system has been evolving since 2013, progressing from classic ML to neural networks to hybrid models with LLMs. The latest upgrade brought the biggest gains in years, with some categories like apparel jumping +40%. Still, this is an incremental improvement, not an overnight transformation.

When Data Caps AI Potential

A good recommendation engine needs a complete picture of user behavior: past visits, searches, purchases, and intent. Without comprehensive data, AI is merely guessing.

Some categories, especially niche or high-complexity ones, don’t generate enough data for solid recommendations. As a result, you have some good, but not great suggestions, leaving plenty of room for improvement.

Fighting Fraud with AI

One big win has been building a multi-layer fraud detection system. Some layers handle simple tasks, catching obvious red flags like suspicious IPs. Others use AI to score behavior patterns and identify likely scammers.

The goal is to disrupt the scammer’s business model. For example, sellers must pre-load funds, and if an account is flagged, it’s blocked before any payout. Make fraud harder on your platform than anywhere else, and most scammers will move on to easier targets.

AI in Moderation and Categorization

Product moderation and categorization used to be manual, messy work. Now AI handles most of it by analyzing both text and images to slot products into the right categories.

Still, accurate categories aren’t just for neatness. They power better recommendations, upsells, and cross-sells. AI’s ability to spot mislabeled listings (whether accidental or intentional) keeps the catalog clean and the shopping experience smooth.

What’s Next for Recommendations

The future might bring proactive AI nudges (“Hey, you bought dog food last month. Ready to reorder?”), but adoption will depend on privacy laws and user comfort.

More likely is the rise in contextual recommendations via chat or voice interfaces. Instead of navigating categories, you might say, “Find me grain-free dog food for a 10-year-old lab,” and the assistant will serve up a shortlist instantly.

Monetizing the AI Search Shift

Google and others aren’t going to walk away from their ad dollars. Expect them to weave sponsored placements into AI-generated answers much like they do with search results today.

The interface will change. The revenue model? Probably not as much as people think. History shows even platforms that initially swear off ads eventually find a way to work them in once the audience is big enough.

Key Takeaways

Running a big product today means keeping a lot of moving parts in sync: dozens of teams, features deploying across multiple platforms, and fast pivots when the tech landscape shifts.

AI is now part of the core toolkit, working behind the scenes to block fraud, organize messy product catalogs, boost search visibility, and serve smarter recommendations. Mobile apps have shifted from nice-to-have to make-or-break, and conversational interfaces are starting to reshape how people shop.

To stay ahead, pull all these pieces into a clear and focused strategy where every shift in tools or trends creates opportunities rather than challenges.