I just finished working on this project! 🚀

MotorCensus is a data-driven vehicle research platform built to help people understand what owners actually report before they buy.

At its core is the U.S. federal vehicle complaint record: the public NHTSA ODI dataset, where car owners across America file complaints about real problems with their vehicles. I built a pipeline that ingests over 2.1 million raw complaint rows, cleans and deduplicates them into 398K matched incidents, and turns them into statistically honest reliability signals for every covered model year.

This one was a fun challenge because it is not just a website. It brings together a large complaint dataset, model-year normalization, signal scoring, filtering, comparison flows, and hundreds of structured pages while keeping everything fast and easy to navigate.

A few technical highlights:

-> 2.1M+ raw NHTSA complaint rows processed into 398K+ deduplicated complaint incidents

-> 731 evidence-qualified model-years, each published with sample sizes and bootstrapped 95% confidence intervals

-> Cohort-relative signal scoring: a 2013 midsize sedan is only ever compared against 2011 to 2015 midsize sedans, never against the whole market

-> Complaint signal scoring and model-year comparisons

-> Dynamic browse and research flows across 800+ statically generated pages

-> SEO-ready architecture with generated sitemaps and production indexing

-> Performance-focused static architecture with a tiny JavaScript footprint, so every page loads fast

The tricky part was making a technically heavy data system feel simple and intuitive for someone who just wants to answer one question:

"Should I actually buy this car?"

Live link: HERE

Really happy with how this one came together.

Looking forward to working with clients here on Contra! 🤝

Cheers, and thanks for reading!