Order.co Takes the Pain Out of Procurement with up to 7x Faster Analytics

Thousands in unapproved spending. Countless hours lost to manual invoicing. It doesn’t matter if you’re a small local coffee shop or a national chain: Procurement and payments can be a pain.

That’s why Order.co exists. The AI-powered procurement and spend management platform helps organizations simplify purchasing while gaining greater visibility into company spending, all in one place.

As Order.co’s business continued to grow, so did its need for reliable analytics, customer-facing data products and machine learning workloads. The team’s existing data warehouse was becoming increasingly difficult to scale.

“Our Redshift environment had development workloads, production pipelines, ad hoc analytics and real-time traffic all competing for the same resources,” says Victor Darmadi, Staff Data Engineer at Order.co. “Ad hoc queries could impact production, creating operational risk.” Expanding its existing infrastructure wasn’t the solution. Adding more clusters would increase infrastructure costs while introducing additional operational complexity.

Rather than continuing to scale its legacy architecture, Order.co selected Snowflake to simplify operations, improve performance and establish a modern data foundation capable of supporting future AI innovation.

The Order.co team started its migration to Snowflake with three clear-eyed business objectives.

First, they needed to eliminate production interruptions caused by resource contention. Second, their new data and AI platform would need to simplify data platform operations and reduce administrative overhead. Third, any sustainable solution would need to improve analytics performance while supporting continued growth.

On its previous platform, separating workloads required maintaining multiple clusters, each adding infrastructure cost and management effort. Engineers also spent significant time tuning workload management queues, priorities and cluster configurations to balance competing workloads.

After migrating to Snowflake’s independent virtual warehouses, these operational constraints largely disappeared.

At the same time, the company saw immediate performance improvements.

“After moving to Snowflake, we saw query performance improve by two to seven times,” Darmadi says. “It immediately reinforced that we had made the right platform decision.”

Snowflake also let Order.co simplify its architecture. The company had been running two separate systems because its warehouse couldn’t efficiently support a number of large data objects. By migrating to Snowflake, Order.co consolidated these workloads onto a single platform while taking advantage of native JSON processing capabilities that simplified development.

Rather than attempting to modernize every workload during migration, Order.co deliberately chose a phased lift-and-shift strategy.

Order.co organized the migration into six waves aligned to the company’s medallion architecture, beginning with bronze and silver data layers before transitioning to gold-layer analytics, dashboards, machine learning inference pipelines and reverse ETL workloads.

Working alongside Snowflake’s Forward Deployed Engineering (FDE) team, Order.co established governance, security and operational best practices while running the migration in parallel.

“We started with Snowflake’s activation checklist covering everything from network policies to resource monitors and governance,” Darmadi says. “That gave us a strong operational foundation from day one.”

The phased approach allowed Order.co to migrate approximately 5 TB of data, support more than 200 Metabase users and complete roughly two months of development work with minimal business disruption.

AI played a central role throughout the migration.

Order.co combined Snowflake AIM migration agents with CoCo and custom migration skills developed by the Forward Deployed Engineering team to accelerate dbt migration, automate dashboard conversion and validate migrated data.

Rather than simply generating code, AI significantly reduced one of the most time-consuming aspects of migration: validating that data remained consistent after every workload moved to Snowflake.

“The AIM migration agent in Snowflake CoCo saved us a tremendous amount of time. Not just generating migration code, but validating data quality by comparing results before and after migration,” Darmadi says.

The AI-assisted approach proved particularly valuable during the migration of nearly 9,000 Metabase dashboards and reports.

"The Forward Deployed Engineering team at Snowflake gave us a strong foundation to build on," Ramaswamy says. "Instead of starting from scratch, we could immediately begin iterating on the migration tooling. That head start saved weeks of work."

These reusable migration assets let a single engineer complete the dashboard migration while continuously refining the automation throughout the project.

Today, Order.co primarily uses Snowflake to power dbt transformations and enterprise analytics across its data platform. While the migration intentionally focused on moving existing workloads before optimization, the company now sees significant opportunities to expand its use of Snowflake.

Near-term priorities include optimizing dbt workloads, improving compute efficiency and building new AI-powered experiences for business users.

"We're seeing growing demand from business teams to interact with data using AI," Darmadi says. "That's why Snowflake CoWork and Cortex Agents are high on our roadmap."

With a modern data and AI platform now in place, Order.co is well positioned to innovate and scale intelligent business applications and additional AI-driven capabilities without needing to redesign its underlying data architecture.

Reflecting on the project, the Order.co team credits a few core choices for its success — and has advice for organizations planning their own data warehouse modernization initiatives. 

Running legacy and new platforms in parallel during its migration reduced business risk, while adopting a lift-and-shift approach first allowed the team to optimize workloads after migration without slowing its transformation down.

Planning warehouse sizing and workload isolation early helped Order.co maximize efficiency, while using AIM to accelerate migration and data validation meant the team could migrate without straining internal resources. Aligning migration timelines with existing data warehouse contract renewals kept the process largely painless.

In the end, Order.co ran its own migration the same way it has redefined procurement for its customers: Consolidate what’s scattered, cut out the busywork and leave people with one place to get everything they need.