FactorPrism®
Case Study

Deep Mining the NYC 311 Dataset

How FactorPrism® analyzed 50+ million records to uncover hidden patterns in New York City's service requests—in under an hour.

50 Million Records, Countless Patterns

New York City's 311 service handles millions of non-emergency calls annually, from noise complaints to pothole reports. With over 50 million records spanning multiple years, the NYC Open Data initiative invited the community to help discover patterns in this vast dataset.

Traditional analysis would require teams of analysts spending months to identify meaningful trends. Even then, subtle patterns and interaction effects would likely remain hidden. We decided to demonstrate how FactorPrism® could surface these insights automatically.

The Approach

Using FactorPrism®'s advanced algorithms, we analyzed the period from September 2013 to March 2017—a timeframe showing clear growth trends. Our goal was to understand whether this growth was uniform across all service types or driven by specific hidden factors.

Key Findings

1. Overall Growth Masking Critical Trends

While 311 usage grew 11% overall during the period, this headline number obscured crucial patterns:

+
Housing Authority (HPD) complaints declined 25% A positive trend completely hidden by overall growth
~
Growth wasn't uniform Concentrated in specific service categories
Seasonality effects varied Amplified in some categories, dampened in others

2. The 2014 Pothole Crisis

FactorPrism® isolated a dramatic spike in pothole complaints in 2014 that was distinct from general street condition issues:

Pothole complaints surged to 3x normal levels in 2014. This spike was independent of other street condition complaints—a pattern that would have been missed by looking at aggregate street complaints. This type of granular insight allows city planners to understand specific infrastructure failures rather than general trends.

3. Hidden Seasonal Patterns in Water System

Water system complaints showed fascinating seasonal patterns with anomalies:

~
Sharp spikes every July and December Like clockwork seasonal stress
Summer 2016: Unusually severe spike Above historical patterns
+
Winter 2016: Unexpectedly mild Suggests infrastructure improvements

The Power of Automatic Pattern Detection

What makes these findings remarkable isn't just their value—it's how they were discovered.

Traditional Approach

  • Team of 3-5 analysts
  • 2-3 months of work
  • Manual hypothesis testing
  • High risk of missing subtle patterns
  • Difficulty isolating interaction effects

FactorPrism® Approach

  • Single analyst
  • Under 1 hour total time
  • Automatic pattern detection
  • Surfaces hidden correlations
  • Isolates pure effects automatically

Technical Excellence

FactorPrism®'s algorithms excel at this type of analysis because they separate overlapping effects (the 25% decline in housing complaints was invisible in aggregate data), identify pure signals (the 2014 pothole spike was isolated from general street trends), and detect anomalies in patterns (the mild winter 2016 water system spike stood out against historical patterns).

What This Means for Your Business

If FactorPrism® can find these needles in NYC's 50-million-record haystack, imagine what it can uncover in your data:

For Retailers

Separate true product performance from category trends, seasonal effects, and marketing impacts

For SaaS Companies

Identify which customer segments are actually churning while overall growth looks healthy

For Operations Teams

Detect when normal patterns break—indicating either problems to fix or improvements to replicate

Ready to Uncover Your Hidden Patterns?

Don't let critical insights stay buried in your data. What would have taken NYC months to discover, FactorPrism® found in under an hour.

Get it on Snowflake Marketplace