WeWork cuts projected costs for data movement by two-thirds with Airbyte

Company Size

2000+ employees

Region

Global (headquartered in New York)

Industry

Real estate, flexible workspace

Sources

Salesforce, PostgreSQL, MySQL, Google Sheets, internal business applications

Destination

Snowflake

Tech Stack

• Source(s): Salesforce, PostgreSQL, MySQL, Google Sheets, custom internal applications
• Destination(s): Snowflake
• ELT: Airbyte
• Orchestration: Apache Airflow, managed by Astronomer
• Transformation: dbt

Key Results

• Reduced data movement costs by two-thirds compared with projections under previous vendor’s new pricing model
• Moved hundreds of connections to Airbyte, retiring legacy pipelines along the way
• Matched existing capabilities like for like, with reliable, set-and-forget syncs
• Consolidated every source, from Salesforce to global building access data, into a single Snowflake warehouse
• Built a reliable data foundation as the team leaned into AI initiatives

About WeWork

WeWork is a leading global real estate platform, purpose-built to empower businesses of all sizes with flexible workplace solutions, hospitality-driven experiences, best-in-class services, and innovative technology. With a curated portfolio of over 600 thoughtfully designed locations, spanning 45M square feet worldwide, WeWork helps more than half a million members — from emerging startups to Fortune 100 companies – perform their best work, today and tomorrow.

Running a global network of buildings generated a steady stream of data: sales activity, property and member systems, and entry and exit events from every location. Turning that data into something the business could use fell to WeWork's data engineering team.

Steven Rowland, Data Engineer at WeWork, along with the wider data engineering team, has been building the company’s platform for years, and it shows.

"Our warehouse, infrastructure, and tooling are mature at this point, and the team is incredibly skilled," Rowland said. "I'm impressed every day by what my teammates can do."

The pursuit of cost efficiency inspired a change

WeWork's data stack had evolved alongside the company.

"Years ago, we ran a homegrown ETL tool called Data Manager," Rowland said. "Eventually we were ready to move on from building it ourselves." WeWork moved most of its ingestion to a new platform, starting when it was still a small company, and we stayed a customer of theirs for many years as both businesses grew.

However, the costs for WeWork to use the company continued to rise throughout the years. For the data team, the decision to make a jump from the old platform came down to cost and convenience.

"Data integration is something you want running quietly in the background without having to think about it. That's the reliability we were looking for when we moved to Airbyte."

A replacement that made sense from the first connection

WeWork evaluated several products before making a decision. "We looked at a number of tools and met with several teams," Rowland said. "Some of the options were intriguing, but we preferred a more mature and polished solution."

Airbyte stood out for how easy it was to reason about. "What drew us to Airbyte was how logical the workflow is," Rowland said.

"You create a source, choose a destination, and connect them. It just makes sense."

What Airbyte gave WeWork:

• A simple, logical setup. The source, destination, and connection model made pipelines easy to understand and quick to configure.
• A managed service with minimal upkeep. Airbyte ran the infrastructure, so the data engineering team didn't have to spend time maintaining ingestion.
• Coverage for every source the business depended on. A number of WeWork’s software programs across the business, including systems that run WeWork's buildings, all consolidated in one layer through one platform. For the few sources without an off-the-shelf connector, the team built its own with Airbyte's Connector Builder.
• Capacity-based pricing. WeWork bought capacity in Data Workers instead of paying for the volume of data it moved, which kept its data integration spend flat instead of tripling.

Cleaning house during the migration

Moving hundreds of jobs off a platform the team had used for years was also a chance to tidy up. "The migration was a good opportunity to clean house," Rowland said. "We consolidated and retired a lot of legacy pipelines that weren't adding much value." By the end of the migration, everything flowed into a single destination.

Same capabilities, a third of the price

The goal of the migration was to solve a cost problem without giving anything up, and on that measure it delivered.

The savings didn't come at the expense of reliability. "The underlying technology is solid, and it's held up," Rowland said. "We've been very pleased with the performance and reliability."

A steady foundation for the AI era

With ingestion running quietly in the background, WeWork's data engineers could put their attention where the business needed it. Like many data teams, their focus is increasingly going towards AI, and Rowland is heavily invested in building the data foundation for these projects.

"We're leaning into AI now, and it's been interesting to watch the focus swing back to the data side of the business," he said.

Reliable, low-maintenance data movement was the result that mattered most. Every source landed where it needed to on schedule, the bill stayed flat, and the data engineering team spent its time building rather than maintaining.

"As a managed service, Airbyte takes very little maintenance on our side, which is exactly what we want," said Rowland. "We set it and forget it."