DAT cuts compute 75% at 15K QPS migrating from Neon to PlanetScale Postgres

This wasn't a migration for fun. We analyzed our traffic patterns and needed a platform that gave us cost predictability and operational control without sacrificing performance. PlanetScale delivered all three.

Ahsan Nabi Dar, Co-founder & CTO, DAT Systems

About DAT

DAT Systems is a Pakistan-based startup that helps organizations manage and monitor field staff operations at scale—including workforces where smartphone literacy and app installs aren't a given.

Rather than asking workers to install yet another app, DAT built on the messaging apps they already use. Field workers send updates—text, photos, videos, voice notes, GPS pins—over WhatsApp, Telegram, and WeChat, and managers monitor everything in real time through enterprise dashboards.

The stack is lean: Elixir on Fly.io, Oban for background jobs, and Postgres as the source of truth.

A relentless, write-heavy pipeline

DAT's database workload is much heavier on writes than a typical CRUD app. Every message from the field is ingested, processed, and audit-logged. Full traceability is a hard requirement for DAT's public sector clients. Here's what their workload looked like:

• ~11 million events per month
• Every inbound message triggers 6–7 background jobs, all persisted in Postgres via Oban
• Every application change writes an audit log entry
• ~15,000 queries per second at peak load

The busiest stretch comes during Eid cleanups, when thousands of workers send updates around the clock for three straight days.

Why DAT left Neon

DAT had been on Neon Postgres since its proof of concept. The team built deliberately on serverless infrastructure to take advantage of autoscaling, scale-to-zero, and branching for dev and test environments. Neon served DAT well through the search for product-market fit and the fast growth that followed.

A few months ago, Neon kindly nudged us: "Hey, you've been running hot for 48+ hours—maybe take a break?" We appreciated that care. Neon was a fantastic partner as we scaled: autoscaling, startup-friendly pricing, and great support. But after their Databricks acquisition, things changed.

Ahsan Nabi Dar

For DAT, three things changed at once:

1. The startup plan it relied on was sunset.
2. DAT was moved to Pay-As-You-Go billing.
3. Its projected costs were estimated at 4× what it had been paying.

Even after downgrading its plan and giving up features it depended on, DAT still faced a 4× price increase. The team spent months evaluating alternatives. By this point, DAT's growth had stabilized and its traffic patterns were well understood, so it no longer needed aggressive preemptive autoscaling to absorb the unknown. It needed predictable pricing and operational control.

Why PlanetScale

PlanetScale had been on Ahsan's radar for a while, and when PlanetScale launched Postgres support the previous year, DAT watched closely. Now, with a stable workload and a clear cost ceiling in mind, the timing was right.

Over a single weekend, DAT migrated its live database from Neon to PlanetScale Postgres, with help from the PlanetScale team on right-sizing the instances.

Like any live database move, it had challenges, especially right-sizing after years of trusting Neon's autoscaling to do the thinking for us. Once we had the actual shape of our workload, the rest was straightforward.

Ahsan Nabi Dar

The results after moving to PlanetScale

DAT's PlanetScale Postgres database now handles ~15,000 queries per second on roughly a quarter of the compute, with double the replicas and a predictable cost line heading into next year's growth.

• ~75% less compute. After right-sizing for its stable workload, DAT now runs on one quarter of the compute Neon's preemptive autoscaling would have charged it for on the same workload. And, in Ahsan's words: "We could go lower."

• High availability by default. PlanetScale Postgres ships with two replicas out of the box at no additional charge.

• Query observability. PlanetScale Insights surfaced resource-heavy queries that DAT fixed shortly after migrating.

• Predictable pricing. DAT's bill is now driven by its workload rather than a pricing model it doesn't control, so it can plan costs around growth.

• More extensions. DAT picked up pg_duckdb, which PlanetScale supports out of the box, for fast analytical queries over its operational data.

• An upgrade to Postgres 18. The migration doubled as a major version upgrade, with no separate maintenance window.

Get started

Sign up to create a PlanetScale Postgres database, or reach out to talk with our team. If you're moving an existing database, we can help you migrate.

Thanks to Ahsan Nabi Dar and the DAT team for sharing their story.