// case studies
Real engagements, real numbers.
A short index of projects where measurement, speed, or custom engineering changed the decision the business was about to make. Each one is written up with what we found, what we built, and what stayed true after the engagement ended.
// montreal e-commerce electronics · marketing measurement
ROAS > 10× Hiding in the Data — While the Dashboard Said “Cut It”
A retailer was about to kill their best-performing ad channel. Linear attribution on observed revenue was hiding the cohort economics. Kaplan-Meier survival curves and cohort-level LTV projection inverted the picture.
Read the case study →// distribution center · dock capacity analysis
When the Average Truck Doesn’t Exist — Validating a New DC’s Dock Door Plan
The architect proposed a door count. A Montreal-based furniture company was skeptical. Hartigan’s Dip Test, a Gaussian Mixture Model, and service-time decomposition revealed LTL trucks consume 2.3× more dock time per unit than FTL. Satellite imagery today confirms the expansion was right.
Read the case study →// sharkninja emea · ERP-triggered BI outage
Custom Dashboard Live in 2 Weeks — While the $4K/Month Tool Stayed Broken for 9 Months
An ERP migration silently broke every BI report. The team was reconciling revenue in Excel every morning — including for the CEO. We were on-site the following week; nine months later the official tool was still broken.
Read the case study →// 100M-row data warehouse · PostgreSQL → BigQuery
100M-Row Data Warehouse in 4 Weeks
A consulting firm quoted $200K+ on an 18-month timeline. We scoped differently, began the following week, and shipped a BigQuery warehouse with 11 dashboards, 3 models, and a marketing analytics platform in four weeks.
Read the case study →// structube · payment reconciliation
Eliminated 4 Hours of Daily Manual Work in 1 Week — Still Running 5 Years Later
A daily reconciliation burning 2–4 hours of finance time, every day. Replaced in a week with an automated pipeline. Five years on, still running without intervention.
Read the case study →Have a measurement or reporting problem you can't get a clean answer to? Tell us what you're trying to figure out →