How Deliverr Managed 2X Data Volume with Near Real-time Replication Using Hevo

Emmet Murphy, Staff Software Engineer at Deliverr saves 2 man-weeks a month by automating their data replication from MySQL to Snowflake using Hevo.

Get a Demo

2X Data Load

Hevo has enabled Deliverr to replicate 2X more data in almost real-time for faster reporting

80 Hours of Saving

The combination of Hevo and Snowflake improved reliability, reduced troubleshooting time and saved 2 man-weeks a month

10% Rise

The data & analytics team at Deliverr increased productivity by 10% using Hevo’s fully-automated data pipelines

Deliverr has a 40 member strong engineering team that consists of growth engineers, product developers, and data engineers. Also, they have a separate data science team of 8 members that consists of data scientists and analysts. Emmet works as a Staff Software Engineer at Deliverr and he heads their data engineering function. He is primarily responsible for creating data flows, deploying machine learning models, managing data warehouses, and ensuring that the entire organization gets access to the most complete and accurate data at the right time.

Emmet Murphy

Staff Software Engineer

Data Challenges

Deliverr was founded in 2017 with a mission to enable sellers, regardless of size, to delight their customers with fast and cost-effective fulfillment. As the world is following the Amazon business model that ensures the same day or next day delivery, Deliverr is helping businesses, marketplaces and sellers match their customer expectations by improving their delivery speed.

Data plays a strategic role in Deliverr’s business model, they leverage proprietary data coming in from a multitude of proprietary sources, ML algorithms, and a network of warehouse partners that help them strategically position inventory and enable fast e-commerce deliveries to customers.

Emmet says, We use a lot of microservices for different business functions and all the microservices data is stored in multiple MySQL instances. We were using FlyData to integrate our MySQL data into Amazon Redshift. But we were facing several challenges with Redshift and FlyData, Redshift wasn’t scaling the way we wanted and at times business teams were not able to get the required reports due to server overload issues. Also, creating data pipelines on FlyData was a manual task, the tool was inefficient to handle schema changes resulting in multiple pipeline failures and due to lack of visibility into pipeline status, we needed to rely on FlyData's customer support to resolve any problems.

After facing all these challenges for several months, the data & analytics team at Deliverr decided to look for an alternative to Redshift and FlyData. Emmet recommended a switch to Snowflake and also started looking for a suitable data pipeline solution.

The Solution

After successfully setting up the data warehouse on Snowflake, the next target for Emmet was to look for alternatives to FlyData. After extensive research, Emmet narrowed down to 4 tools - Fivetran, Stitchdata, Matillion, and Hevo. After trying them for almost a month, he finally decided to go ahead with Hevo because of its data quality, auto-mapping feature, fully automated data pipelines, and ease of usability.

Emmet says, Getting onboarded to the Hevo platform was super easy. I specifically liked the dashboard feature that provides a real-time update on data load status and latency. Hevo’s support is exceptional and we were extremely happy with this product because it exceeded all our expectations.

Earlier Emmet and the team were spending 1 or 2 days a month troubleshooting Redshift issues, but now, Snowflake is always up and running smoothly. Also, with Hevo, Emmet and the team streamlined their data replication process, improved the quality of their reports and now they hardly face any problems with replication delays.

The bulk of our reporting is set up on Metabase and it is accessed by almost all our employees, partners, and sellers to view important business KPIs. Our operations team uses reports & ML models to manage the warehouse network, monitor on-time delivery, and ensure seller satisfaction. And Hevo plays a crucial role in providing the most accurate, real-time, and unified data to all our business stakeholders, says Emmet.

Emmet and team are using MySQL BinLog replication as a source and use over 25 different microservices pipelines on Hevo that load their multi-functional data into Snowflake. Their reporting tool (Metabase) sits on top of Snowflake and provides near real-time data for their management and operations team to track the delivery network status.

Hevo Deliverr Data Stack

Emmet says, Hevo helps us centralize our data to Snowflake and effectively use it for reporting. Deliverr’s product and engineering teams manage their own set of microservices, each team creates Hevo data pipelines on its own, without needing help from data engineers or DevOps.

Key Results

In the last 6 months, since Deliverr switched to Hevo and Snowflake, their data volume more than doubled, query latency improved by around 25-40%, data warehouse reliability has improved from around 97% to 99.98%, and replication lag reliability has improved from around 97% to around 99.8%. Also, Emmet and team were able to save around 1 man-week a month by just automating their data pipelines using Hevo and another man-week a month for troubleshooting gains with the combination of Snowflake and Hevo reliability improvements.

One of the biggest reasons why I would recommend Hevo is because of its lowest price-performance ratio as compared to the competition. It is definitely one of the best solutions if we take into consideration 3 major aspects - scalability, productivity, and reliability, says Emmet.

In the last year alone, Deliverr’s customer base has increased by 11X, warehouse footprint by 4X, onboarded thousands of merchants and over 50% of people in the U.S. now live within 100 miles of their fulfillment centers. Deliverr is truly delivering a multi-fold growth and has recently raised $40M in Series C funding. They are planning to use this funding to advance their machine learning capabilities so that they can predict demand and stock warehouses accordingly. Hevo is extremely proud to be a part of their growth journey and looking forward to fueling the most accurate, real-time, and unified data for strengthening their ML capabilities.

Excited to see Hevo in action and understand how a modern data stack can help your business grow? Sign up for our 14-day free trial or register for a personalized demo with our product expert.