As travel patterns shifted post-pandemic, flexibility became a top priority for guests. People began searching for stays that fit their evolving lifestyles whether for work, longer trips, or remote living. To support this shift, the team needed a way to visually classify and enrich a massive volume of property listings. But tagging millions of images across global inventory proved too time-consuming and complex using traditional tools.
They adopted Databrewery’s Annotate and Catalog tools to speed up the process. With smarter workflows and powerful visual search, the team was able to organize, label, and categorize property photos at scale. Databrewery made it easy to surface relevant attributes, create structured data from images, and handle the constant influx of new listings efficiently.
They adopted Databrewery’s Annotate to speed up the process. With smarter workflows and powerful visual search, the team was able to organize, label, and categorize property photos at scale. Databrewery made it easy to surface relevant attributes, create structured data from images, and handle the constant influx of new listings efficiently.
To keep up with rapid growth and support smarter decision-making across the business, this team needed to automate and standardize their machine learning workflows. With various departments working with different formats of unstructured data, especially listing images it was critical to find a scalable way to classify, enrich, and manage that data. Without a consistent approach, running data science initiatives across teams was inefficient and fragmented.
To reduce the time and cost of manual labeling, they began building fully automated ML pipelines using active learning. These pipelines were integrated directly into their annotation process, so the system could prioritize the highest-impact data points for review. Whether it was operations, customer support, or pricing teams, everyone now works from a shared infrastructure with a unified ontology allowing consistent labeling across use cases.
Using Databrewery’s collaborative features, including real-time comments and issue tracking, the team streamlined review cycles and drastically improved how fast they could create structured, enriched listings with relevant visual and contextual metadata. Annotators, reviewers, and model trainers now work together more fluidly reducing friction and improving the end result.
Just three months into the rollout, the company’s pipelines are largely automated. Most annotations are generated by models and verified by subject matter experts. Human labeling costs have dropped significantly, and more than nine million annotation tasks have been completed, a major milestone that reflects both scale and efficiency.