Emora Health: 95%+ Claims Acceptance with DuckDB & MotherDuck

I didn’t want to build an internal analytics platform. Our job is to improve pediatric access and outcomes, not to reinvent infrastructure. We wanted DuckDB-level speed with a managed experience.

Robin Anil

Chief Technology Officer

Emora Health provides virtual mental health care for kids and families with a privacy-first analytics stack powered by MotherDuck — supporting sub-second insights, same-day claims filing, and 95%+ acceptance rates, while keeping sensitive family data secure.

Emora Health provides virtual mental health care for children and teens. That means every operational detail is connected: clinical outcomes, scheduling, eligibility, claims, reimbursements, and the family’s experience across the entire journey.

After migrating to MotherDuck, Emora built a lean analytics foundation that powers the business end-to-end and changed what was possible operationally:

Same-day claims filing for 80-90% of appointments (previously took up to a month)
95%+ claims acceptance rate through rapid iteration on rejection patterns
Sub-second query performance on healthcare data that other warehouses took multiple seconds to process
In the end, this wasn’t just about analytics. It was about thousands of families getting care on time — without having to chase insurance or live in uncertainty.

“This isn’t big data. It’s healthcare complexity for over 120 million families in the US, We have a wide set of systems that all describe the same family journey. The goal is to unify it and move quickly, because speed and privacy matters to families."

Emora Health’s Mission: Shrinking the Pediatric Access Gap

Half of all mental health issues manifest by age 14, yet fewer than 50% of children can access care. In rural America and suburbia, pediatric psychologists are scarce. Emora was founded to close that gap through telehealth, but pediatric care can be harder to find than adult therapy. Parents must approve treatment. Divorced households add complexity. Multiple caregivers mean multiple relationships to manage

"A lot of people told us not to take this on because it’s hard." says Robin Anil, Emora's CTO and former head of research technology at Bridgewater Associates. "But when you understand the need, ‘hard’ isn’t a deterrent, it's the point. So we said, let's go solve it."

Small Data, Wide Problem

Emora started with running analytics on Google Sheets. The founders were meticulous about tracking each metric — one came from finance, the other from healthcare — but spreadsheets couldn't scale.

When Robin joined, he evaluated Snowflake, Databricks, and the usual suspects. The more he dug in, the more he realized: this wasn't a big data problem. It was a small data problem spread across many dimensions — marketing, CRM, EHR, claims, eligibility, and their own Postgres database. Every query needed to marry data across all of them to create a unified view of a client journey.

"I ran a few queries, and it was taking a second to multiple seconds," Robin recalls. "And I look at it — this is small data. This should not take this long. I know this is going to take two milliseconds if I hand-write this thing."

He had used DuckDB in his prior role, where it delivered 10x better query latencies than Trino on their hundred-petabyte lake. When he discovered MotherDuck, the managed version, the decision was easy.

"I didn’t want to build an internal analytics platform. Our job is to improve pediatric access and outcomes not to reinvent infrastructure. We wanted DuckDB-level speed with a managed experience."

The Architecture

Emora's data infrastructure is deliberately lean:

Ingestion: A Lambda function runs hourly incremental loads — built in 24 hours, running reliably for four months
Sources: EHR data arrives as Parquet in S3; other sources (HubSpot, Postgres) follow the same pattern
Processing: Change data loads into staging tables, then merges into main tables by ID
Analytics: The entire company uses Hex for visualization, with MotherDuck as the backend

"We didn't end up using any of the data flow pipelines," Robin explains. "We kept the pipeline deliberately simple: incremental loads, staging tables, and merges into core models. It’s not flashy — but it’s reliable, fast, and easy to evolve. The only time it failed in ten months was a five-minute transient RPC error on MotherDuck's side."

What Changed

The performance gains enabled a fundamentally different approach to claims processing. Before, claims submission and processing took several weeks to months. Parents waited a long time for clarity on costs. Robin recalls a call with a parent who was struggling financially and wanted predictability. "It was a reminder that the U.S. healthcare system pushes uncertainty onto families. We decided to change what we could control."

That conversation was a kick in the pants. The team used analytics built using MotherDuck to rapidly iterate through rejection patterns — analyzing failures, codifying insights back into the pipeline, tracking improvements. Claims filing dropped from monthly batches to same-day for 80-90% of appointments. Acceptance rates climbed to 95%+, which is considered excellent in the industry.

"We were able to rapidly iterate through the cases for failures, mistakes people make, analyze it, come up with insights, and codify it back into our pipeline. And our claims team said, 'We're hitting 95% plus on acceptance right now.'"

The impact extends beyond claims. Every team member now explores data directly in Hex — slicing, dicing, asking new questions.

"Data is all about asking the right question and moving quickly," Robin says. "The win isn’t dashboards — it’s speed. When anyone can explore and answer questions quickly, the organization gets smarter every week."