← Back

How RAD got 2.5 hours down to 2 minutes

A Shopify brand still transforming data on a laptop for two and a half hours — until the pipeline ran in two minutes, without anyone kicking it.

R.A.D data stack
Model
Performance & lifestyle
Channel
Shopify
Reporting
Power BI
Timeline
Four weeks
Infra
Under $20 / month

Meet RAD

RAD sells performance and lifestyle products through Shopify. Power BI was already how marketing, ops, and commercial saw the business. The reports were not the gap. The wait was — and whether anyone trusted the numbers when they finally landed.

The challenge

The laptop was the pipeline

The ingredients of a stack were there. The work still ran on someone's machine. As volume grew, transformations stretched to two and a half hours. Power BI lagged behind the day. A failed run had nothing between it and the reporting tables.

Not “optimise the SQL.” Rebuild the pipeline they already had — so it ran without a laptop, and without yesterday’s numbers.

The solution

Four weeks. Fivetran already pulled Shopify. The job was what happened next. BigQuery and dbt went in layers — Staging, then Mart — so Power BI stopped waiting on a local transform. Compute Engine took the run off individual machines. Tests sat between transformation and the reporting tables; Slack if something broke; snapshots if it needed unwinding. I owned the transformation and reporting side. I worked with Sharaf on the engineering so the model Power BI already used was the one the new stack produced.

  • Move the transform off the laptop onto Compute Engine — prove the wait could go
  • Layer BigQuery and dbt, Staging into Mart, so Power BI read a trusted table
  • Tests, Slack alerts, and backups so a failed run could not overwrite reporting

Shopify into Fivetran, BigQuery and dbt, Power BI on the Mart — Compute Engine, GitHub, tests, Slack, backups around it.

Before and after: 2.5 hours to 2 minutesNo more manual refreshes — the pipeline on a cloud serverStack cost under $20 a monthTests, daily backups, and Slack alertsThe big wins: faster insights, trusted reporting, automated pipeline

The results

  • Transformations from 2.5 hours to 2 minutes
  • Pipeline runs without a laptop refresh
  • Failed runs cannot overwrite reporting tables
  • Power BI no longer waiting on yesterday
  • Infrastructure still under $20 a month
  • History to unwind a bad run

The business could grow Shopify volume without the data stack becoming the wait.

The run happened without someone kicking it. The numbers that landed were the ones they could use. Power BI stopped lagging the day.