The Challenge
As a two-sided marketplace, Wunderflats’ success depends entirely on how efficiently supply and demand are connected. While the data infrastructure had matured significantly under a dedicated BI team, Machine Learning was still uncharted territory. The challenge was not experimentation for its own sake, but turning ML into a reliable, value-generating product capability without losing focus or momentum. Three challenges stood at the center:
- Marketplace discoverability at scale: As inventory and user demand grew, surfacing the right listing to the right tenant at the right moment became increasingly complex. Rule-based ranking could not adapt to shifting user behavior, seasonal dynamics, or the heterogeneous nature of available supply — limiting match efficiency and marketplace liquidity.
- From analytics maturity to ML readiness: Despite a strong BI setup, Wunderflats lacked a clear blueprint for introducing Machine Learning into production. Without a structured assessment of data, infrastructure, and governance requirements, premature ML development risked becoming costly and difficult to scale effectively.
- High-impact use case prioritization: Machine Learning offered many potential applications across the platform, but not all would deliver immediate or meaningful business value. The key challenge was identifying a first use case that directly impacted core marketplace performance and could justify sustained ML investment going forward.
Our Approach
To ensure Machine Learning delivered real marketplace impact from the start, we partnered closely with Wunderflats’ BI and product teams to move from strategy to production in a structured and pragmatic way.
Design Sprint as Strategic Foundation
We started with a comprehensive design sprint to assess data maturity, infrastructure readiness, and business opportunities across the platform. This allowed us to define a clear ML roadmap, align on realistic ambitions, and avoid premature hiring or ad-hoc experimentation — giving Wunderflats the confidence to invest in ML the right way from the outset.
Ranking Engine as Core Marketplace Lever
Based on the design sprint findings, we jointly selected the ranking engine as the first ML use case. As the central mechanism connecting supply and demand, improving ranking quality directly increased booking probability, reduced time-to-contract, and improved satisfaction for both tenants and landlords — making it the highest-leverage starting point available.
Production-Ready Implementation in Phases
We designed and implemented the initial ML infrastructure to support iterative model development and deployment. The ranking engine was delivered in phased releases, ensuring early business impact within a six-month period while creating the technical foundation for increasingly complex models over time.
Embedded Collaboration with the BI Team
Throughout the process, we worked in close collaboration with Wunderflats’ internal teams. By consistently aligning technical decisions with business outcomes and ROI expectations, ML became a shared capability embedded in how Wunderflats builds — rather than an external black box delivered and left behind.
