Over the past decade, alternative data evolved from an exotic research curiosity into a multi-billion dollar cornerstone of institutional investing. As traditional price and volume data became commoditized, hedge funds turned to alternative datasets to gain an informational edge.
However, extracting persistent trading alpha from alternative datasets is notoriously difficult. Many promising signals fail completely when deployed in live markets.
The main categories of alternative data
Alternative data encompasses any dataset outside traditional market quotes and standard regulatory filings:
- Consumer transaction data: Anonymized credit card and debit card receipt feeds tracking real-time retail and restaurant revenue weeks before quarterly earnings reports.
- Geolocation data: Mobile device pings estimating foot traffic at retail locations, manufacturing plants, and distribution centers.
- Web scraping and digital exhaust: Product pricing histories, job opening postings, employee review sentiment, and website traffic statistics.
- Supply chain and shipping manifests: Customs filings, bill of lading documents, and satellite vessel tracking monitoring global container flows.
The three deadly traps in alternative data research
Researchers often discover what looks like exceptional historical performance, only to realize the backtest was distorted by subtle data flaws:
- Look-ahead and point-in-time bias: Historical data files often record when an event happened, not when the data was actually delivered to clients. If a credit card aggregator delivered data on a two-day delay, your backtest must strictly reflect that delivery lag.
- Survivorship and coverage bias: Alternative data vendors frequently improve their vendor coverage over time. A dataset that tracked 200 retail stores in 2018 might track 2,000 stores today. If you evaluate past years using modern coverage assumptions, your results will be heavily distorted.
- Signal decay: Once a dataset is licensed to dozens of major hedge funds, the market impact of their collective trades rapidly erodes the alpha. The most valuable datasets are proprietary, exclusive, or technically difficult to clean.
How to showcase alternative data on your resume
If you want to impress quant research recruiters with a personal project, avoid buying expensive commercial feeds. Instead, build your own automated pipeline:
Scrape public pricing or job vacancy data across an entire industry sector daily. Clean the messy text, handle missing entries gracefully, and verify that your historical snapshots strictly preserve point-in-time integrity. Presenting a disciplined, end-to-end dataset with documented cleaning methodologies will demonstrate the exact qualities top research desks demand.