ParadeDB Powers Modern Treasury's Core UI and Search APIs
ParadeDB powers Modern Treasury's core UI and search APIs, eliminating query timeouts and increasing write speeds by 3x.
Overview
Modern Treasury is the payments operations platform for businesses that move money at scale. The platform offers APIs, dashboards, and workflow tooling to manage the full lifecycle of payments — from initiation and approval through reconciliation and reporting. Modern Treasury's customers span financial services, fintech, and enterprise, relying on the platform to process high volumes of transactions with accuracy down to the penny.
The Problem
As Modern Treasury's customer base grew, so did the complexity of powering its core product UIs. The platform's flexible filter interfaces allow customers to search, sort, and aggregate across payments, transactions, and ledger entries using any combination of dozens of attributes. With many customers storing over 100M records, the permutations of possible query plans made it impossible to cover every case with traditional B-tree indexes.
Our Top K in Postgres blog post explores why B-tree indexes don't generalize well to workloads that combine many filters with flexible sorting.
The team endlessly added B-tree indexes to cover different customer search patterns, but there was no reasonable way to index every permutation without causing write performance degradation. As a workaround, Modern Treasury built summary roll-up tables to support basic aggregations — a common pattern for workloads where Postgres struggles at high record counts. However, these roll-up tables introduced data delays and added significant complexity to the system.
The result was a long tail of edge-case query timeouts. Approximately 5% of queries failed entirely due to unoptimized query plans, and the engineering team was stuck in a reactive cycle of adding indexes to patch individual failures.
The Solution
Modern Treasury needed a backend that could power subsecond responses across its core UIs without requiring a different index for every query pattern. The requirements boiled down to filtering records based on any of dozens of attributes, sorting by any attribute, and returning the top results. Crucially, the system had to be accurate to the penny.
Modern Treasury needed the performance of a search and analytics engine alongside the accuracy of a database.
Matt Higgins, Head of Engineering
The team evaluated Elasticsearch but concluded that the data consistency challenges and deployment complexity it would introduce were not worth the trade-off. While Elasticsearch is mature and feature-rich, the value of Postgres replication and SQL-native infrastructure offered by ParadeDB — as a native Postgres extension — outweighed the bells and whistles.
Built on the solid foundation of Postgres, ParadeDB provides the good parts of Elasticsearch we actually need, without the infrastructure and hassle of data synchronization.
Matt Higgins, Head of Engineering
Modern Treasury deployed ParadeDB as a logical replica of their primary Postgres across 7 clusters. This meant the team could leverage its existing Postgres knowledge and infrastructure. No new data pipelines, no sync jobs, and no new query language. Implementation required only minor adjustments to existing SQL queries.
Results
ParadeDB resolved an entire class of edge-case query timeouts for Modern Treasury. The team no longer needs to think about the many permutations of query plans that previously caused failures. The 5% of queries that previously timed out are now subsecond.
ParadeDB's performance on counting and summing operations enabled Modern Treasury to add real-time counts and aggregations to user interfaces that previously could not support them. The platform now powers grouped aggregations and transaction-to-payment reconciliation workflows that were previously too expensive to run inline.
Perhaps most significantly, ParadeDB allowed Modern Treasury to drop a large number of legacy B-tree indexes that the UI no longer depended on. With fewer indexes competing for write throughput, raw write speeds increased by 2-3x on the platform's most important database tables.
Modern Treasury now uses ParadeDB across its payments, transactions, and ledger products, servicing both customer-facing search APIs and internal analytic processes.
ParadeDB
ParadeDB is the simplest way to add Elastic-quality search to your Postgres. If you have a use case around large scale search in Postgres, we'd love to chat.