Elixir Behind The RealReal: The Software Before the Sale

The RealReal is a marketplace for secondhand luxury goods. Before an item reaches a shopper, it needs to be received, authenticated, photographed, priced and connected to its seller. Every handoff creates information that the next person needs to do their job.
Elixir has been part of the software supporting their business for years. This article explores how The RealReal’s engineers have used Elixir and Phoenix, from seller onboarding and authentication tools to shared live updates. We’ll look at the problems they tackled, the choices they made and what other product teams can learn.
Every item brings its own work
A used bag carries its own condition, history and price. Those details need to stay attached to the correct physical item as it moves between people and departments.
The RealReal describes authentication centers combining luxury experts with technology, alongside receiving, photography, storage and fulfillment. For engineers, that creates a coordination problem. A customer sees a listing, while employees need records that help them move the item toward that listing. Inside the authentication centers
A delayed photograph or an incomplete assessment can hold up the next step. These are examples of the operational dependencies a resale platform needs to manage.
Where Elixir entered the business
In an October 2017 account, a former engineering leader at The RealReal described coordinating a move of the seller acquisition flow from Rails to Phoenix, Elixir’s web framework. He also reported that the new user experience went live that October. Bull’s firsthand account
By June 2021, their engineering manager was recruiting developers to build tools for luxury authentication and counterfeit detection, combining Elixir with machine learning. He described work on better workflows and making data available for decisions. That places Elixir close to the employees handling merchandise and its records. Feyh’s description of the work
More recent public work includes the company’s Elixir service-connection libraries. Its conn_grpc package was updated in March 2026. Package history
Why its engineers reached for Phoenix
A particularly clear explanation comes from their senior software engineer’s August 2021 article about a small team tool.
Her remote colleagues used icebreaker questions to get to know each other. Tracking which questions had already been asked became cumbersome. She wanted a single page where several people could make changes and see updates immediately, while writing the application in Elixir.
She chose Phoenix LiveView. It let the server manage interactions and update the browser, meeting her goal of building the interface without custom JavaScript. Existing familiarity mattered too, as she wanted the project to use technology relevant to her everyday work. Rodante’s published reasoning
The example shows a concrete reason an engineer selected the tools, with requirements readers can recognize in their own internal applications.
What happened in practice
Rodante’s tutorial connected saved questions to live browser updates. A database stored the records, and Phoenix PubSub announced changes so other open pages could reload the relevant information. Her finished example let teammates add questions, mark them as asked and see changes across sessions.
The company’s open-source work exposes another practical concern. conn_grpc manages persistent connections between services using gRPC, a protocol for service calls. Its documentation offers either one connection or a pool that distributes calls across several. It explicitly warns that a single connection can become a bottleneck under load. Connection management documentation
That is a useful trade-off for teams considering Elixir. The runtime supports concurrency, but shared resources still need capacity planning. More simultaneous work can put more pressure on a connection, a database or another service.
Lessons from the inventory rules
The RealReal’s vendor API makes coordination tangible. When an item sells, its available quantity decreases automatically. When an order is canceled, quantity requires a separate adjustment. An approved item also has different editing rules from a draft. Those are published integration rules, and they apply regardless of the service’s implementation language. Vendor inventory rules
For another marketplace, the lesson is to define transitions precisely. What does “available” mean? Who can change it? What should happen after cancellation?
A team might require a physical stock check before making an item available again. Its product rules should say so, and the software should make the next action obvious to an operator. A successful API request is one part of completing that work.
Consider an illustrative photography workflow. An employee records that images are ready, the database saves that change, and colleagues’ screens receive an update. If notification delivery fails, reopening the item should recover its saved status. If uploading fails, the record should show that work remains pending. This is a possible design for another team, drawing on the distinction between stored records and live updates.
Teams operating Rails and Elixir together also need clear ownership of records and rules. Our guide explores practical integration choices and the operational work they introduce. Ruby on Rails and Elixir together
The next story
This article continues our overview of 20 major companies using Elixir. Next, we’ll explore Multiverse, where Elixir helped build a platform for upskilling, workplace learning and career development.
If your marketplace has growing workflows, integrations or live interfaces, Elixirator can help you assess the architecture and build the next part. Tell us where your team is getting stuck. Talk to Elixirator
Frequently Asked Questions
Did Elixir replace all of The RealReal’s Ruby code?
The company’s engineering blog describes using both Elixir and Ruby. Bull’s 2017 account concerns a particular seller flow. The engineering team’s stack description
Does Elixir authenticate luxury goods?
The RealReal’s experts assess goods, supported by software. Feyh described engineers building authentication and counterfeit-detection tools using Elixir and machine learning.
What can another product team learn from this?
Start with a specific workflow, define its saved state and make changes visible to the people doing the work. Evaluate connection limits and recovery behavior alongside the user interface.

