Own Core Advantage, Outsource Rest

Diving deeper into

Andrew Yates, CEO of Promoted.ai, on driving marketplace ARPU with personalization

Interview
If it's all of them, why wouldn't you just be outcompeted by someone who focuses on one and outsources the other pieces?
Analyzed 4 sources

The core risk is organizational dilution, not just technical complexity. A marketplace that tries to become world class at logistics, fulfillment, search ranking, recommendations, seller tooling, and ads at the same time is spreading its best people across multiple hard businesses. The evidence from large platforms is that search and ads alone can each require hundreds of engineers, deep measurement systems, and years of iteration before they drive sustainable ARPU gains instead of clutter and weak ROI.

  • Promoted frames search, recommendations, promotions, and ads as one optimization stack because the same system decides what a user sees on a small screen, measures whether it led to a valuable action, and adjusts ranking or pricing over time. Outsourcing isolated pieces breaks that feedback loop.
  • The practical alternative is specialization. If a marketplace's edge is fast delivery, unique supply, or seller relationships, then spending years recreating Facebook or Amazon style discovery infrastructure can turn into a side quest that absorbs senior ML and infra talent without becoming a durable advantage.
  • The market has already split this work into specialist layers. Mirakl sells the marketplace operating system, then adds Mirakl Ads as a separate monetization layer for retailers that want seller funded placements, showing how even large commerce platforms increasingly buy critical infrastructure rather than build every component from scratch.

Going forward, the winners are likely to be marketplaces that keep owning the piece that makes them unique, then plug in specialist infrastructure for the rest. As discovery and retail media become more important profit drivers, the gap should widen between companies that compound value through better matching and those that pile on disconnected tools and heavier headcount.