EVANA improves accuracy at scale with MongoDB

THEIR CHALLENGE

Unifying disconnected data

Behind every real estate transaction, there is copious paperwork. For many buyers, purchasing a house is something they only need to navigate a couple of times in life. However, for professional real estate investors, each deal involves vast tranches of documents to work through. And they are never an easy read. From contracts and architectural surveys to environmental reports and tax documents, extracting the right insights from these documents can be time-consuming.

That’s why German company, EVANA, is on a mission to digitize property management. EVANA gives property owners deeper insight into all documents relating to their buildings. Additionally, they can create a digital transaction data room for each asset, which makes all relevant documents easily accessible to approved buyers during the sales process.

Using AI and optical character recognition (OCR), the platform analyzes documents, classifies them, allows information to be extracted, and data is automatically sorted into data rooms. This makes it easy for customers to query their data to find the right insights. They can also generate captions for images automatically and use powerful AI to translate a document or create a summary of hundreds of documents.

However, as AI developed rapidly, EVANA was mired in legacy technology eight years after its 2015 launch. 

“We had lots of teams all doing their own thing, tech sprawl, and far too many databases,” says Marc Arndt, Vice President of Engineering at EVANA. “From PostgreSQL to Azure SQL, Elastic Search and MongoDB, data was everywhere and inconsistent. We were also relying heavily on PostgreSQL for AI, but that meant replicating data sets at ever-increasing costs.”

The environment was not just challenging for EVANA to manage, it was risking the customer experience. As Martin Teiber, the current CEO and former Vice President of Customer Success reveals, “Performance issues and instability meant that my team and I were spending a lot of time managing customer expectations and giving estimates for when systems would be fixed.”

This was a far cry from the ideal EVANA customer experience, where Teiber describes having “a stable, performant platform helping them to get their documents sorted, classified, and queryable as quickly and accurately as possible.”

OUR SOLUTION

Improving accuracy and scale

EVANA decided to modernize its environment in 2023. The aim was to establish a single source of truth and eliminate excessive data replication. After careful evaluation of multiple data platforms, MongoDB Atlas was selected, and the company began a phased decommission of its PostgreSQL and other legacy databases.

“The three deciding factors were operational simplicity, flexibility, and developer efficiency,” recalls Arndt. “MongoDB Atlas streamlines database management, which minimizes DevOps overheads. It enables rapid iteration and experimentation, and its unified query language supports efficient development.”

MongoDB Atlas is also well-suited for EVANA’s primary use case—ingesting vast amounts of both structured and unstructured data and applying AI models to categorize it and retrieve relevant insights. The platform is document-based with a flexible schema. This has proved better suited to handle multiple types of data than SQL databases.

Using MongoDB Atlas Search, customers can run natural language queries on their data. For example, they could type ‘summarize all my tenants‘ or ‘find all heating maintenance reports’ and the platform would automatically direct them to the right documents. They can also create a collection of documents relating to one of their properties.

The implementation went smoothly and was handled in-house in around six months. As Arndt explains, “MongoDB is extremely well documented, and the team supported us when we needed them to validate our ideas around the right way to implement the platform. MongoDB’s support team also handled our tickets really quickly.”

For small companies, managing change can be challenging. For EVANA, staff were quick to adapt; however, customer reassurance was vital. 

“Our customers were wary of change and feared platform instability. Previously, we only released every two months and sometimes deploying new features would break other things,” says Teiber. “We spoke to them to explain what we were doing and that we’d be able to deploy new features much faster—with up to eight releases a day. Soon, they saw for themselves that there was nothing to worry about, and our challenges were in the past.”

Delivering a positive experience to developers is essential to EVANA’s culture. It attracts motivated self-starters who love working independently to produce high-quality code. MongoDB’s query language unlocks a wide range of capabilities. These range from CRUD operations to advanced pipelines, vectorization, and graph queries. All of which reduces the need for developers to learn multiple database syntaxes.

EVANA has enhanced security using OpenID Connect (OIDC) for authentication. This removes the need for passwords and reduces the risk of credential leaks. “Both human and machine identities are securely managed,” added Arndt.

With MongoDB Atlas, EVANA is always running the latest version of the platform. Less time is spent on updates and maintenance. Instead, the team can run more pilots, test, and deploy new features rapidly, confident that the system will alert them to any performance issues. They also don’t need to set up new schemas or tables for new use cases.

OUTCOME

A better experience for everyone

Over two years, EVANA has transformed and modernized its customer-facing platform. This has given real estate managers a reliable way to extract information from vast volumes of data, to generate captions, and categorize documents. Accuracy has improved as the volume of documents has increased from 6,000 to 30,000 per hour, reaching 95% accuracy over 38 million scanned pages.

This will help customers with due diligence during a property sale and simplify extraction of the most crucial information—structural issues noted in planning documents or clauses within contracts, for example. Previously a manual process, this will help save thousands of hours of reading. In the future, customers will be able to use AI to query relevant data further and receive an immediate, accurate response.

“With MongoDB, we’ve evolved into EVANA 2.0. It helps us keep up with the rate of change and deliver a great customer experience,” says Teiber. “We can also listen to their feedback and deliver changes the very next day if they have a great idea for the platform.”

Centralizing on MongoDB has eliminated instability and performance issues, creating a single source of truth with no duplication. It’s easier to manage and more cost effective. One DevOps resource monitors performance and metrics, and staff no longer need to focus on backups and patching.

“We’ve streamlined operations, sped up deployment cycles and reduced IT overheads,” explains Arndt. “The design enables us to iterate quickly so we can focus on experimentation and validation rather than planning.”

Next, EVANA is working to make the functionality it has developed available to other industries. As AI is language-agnostic, EVANA can expand into new markets without worrying about language barriers. The principle of extracting meaningful insights from vast amounts of data can be applied to medical, legal, and government use cases.

All these gains have enhanced EVANA’s customer experience and improved its competitive edge. MongoDB also has all the capabilities to support the team’s AI ambitions. As Arndt concludes: “If you want a database that just works and makes your life easier, MongoDB is the best choice now and for the future.”

Developers have also created data rooms for customers to use during transactions. This gives customers the power to move packets of documents into a secure data room to share with property buyers. These documents are queryable, enabling buyers to locate relevant details rapidly.

Meanwhile, new structural functionality enables customers to create collections of up to millions of documents and use MongoDB Atlas Search to query them and ask follow-up questions based on the results. The flexibility of MongoDB Atlas also supports the transaction feature of EVANA’s platform, which enables customers to make all relevant documents available to prospective buyers at the click of a button for each property.