Using Alternative Data and Channel Checks Responsibly

The practical point: alternative data can give you a 2-6 week edge on quarterly results, but only if you triangulate it with fundamentals, respect legal boundaries, and accept that 30-50% of alt-data signals are noise that fails to replicate out of sample.

Why Alternative Data Matters (and Why It Can Mislead)

Alternative data refers to non-traditional information sources--web traffic, app downloads, credit card panels, satellite imagery--used to estimate company performance before official filings. Hedge funds spent an estimated $1.7 billion on alternative data in 2020, growing 20-25% annually since.¹

The point is: you are not buying "insight"; you are buying partial, noisy signals that require calibration. A single alt-data source might explain 15-40% of revenue variance for e-commerce, but only 5-10% for B2B software with enterprise contracts.

Types of Alternative Data (What You Can Access)

Web Traffic Data (SimilarWeb, Semrush): Estimates monthly visits, time-on-site, and page views. Works best where online engagement directly correlates with revenue. Key signals: month-over-month visit growth, bounce rate changes, geographic shifts. Limitation: estimates can be 15-30% off versus actual server data.

App Download and Usage Data (Sensor Tower, data.ai): Tracks downloads, DAU/MAU, session duration. The point is: downloads measure acquisition; DAU/MAU measures retention--and retention drives revenue. A >20% DAU/MAU ratio is strong for consumer apps. Limitation: install estimates can be 20-40% off for smaller apps.

Credit Card Panel Data (Second Measure, Earnest): Aggregates anonymized transactions from panels representing 2-5% of U.S. consumers. Provides direct revenue proxy for consumer-facing companies. Key signals: spend growth vs. prior year, customer count vs. spend per customer. Limitation: panels skew by bank and demographics; B2B revenue invisible.

Satellite and Geolocation Data: Tracks parking lot traffic (retail), oil storage levels (energy), construction activity (real estate). Key signals: car counts vs. prior quarter, storage fill rates. Limitation: weather and local events create noise; interpretation requires domain expertise.

Legal and Ethical Boundaries (The Line You Do Not Cross)

The point is: legal alternative data comes from observable public behavior or consensual data sharing; illegal data comes from stolen information, hacking, or material non-public sources.

What Is Legal:

• Web scraping of publicly accessible pages (check terms of service)
• Aggregated, anonymized transaction data from panels with user consent
• Satellite imagery of public spaces
• Job posting analysis from public job boards

What Crosses the Line:

• Corporate espionage: obtaining internal documents, hacking systems
• Tipping: receiving material non-public information from insiders
• PII exposure: using data that identifies individuals without consent

Why this matters: the SEC has prosecuted cases where expert network consultants provided material non-public information disguised as "industry insights." If your data source cannot be explained publicly, you have a problem.

Channel Check Methodology (Primary Research Done Right)

A channel check is direct contact with suppliers, distributors, or customers to assess real-time performance. This is legal when done correctly.

The Right Way:

1. Identify yourself and your purpose. You are an investor researching the industry.
2. Ask about industry trends, not company secrets. "How is demand?" not "What did Company X order?"
3. Aggregate multiple sources. One distributor is anecdote; five are data.
4. Document your process. Written notes of calls, dates, topics.

The Wrong Way: Pretending to be a customer, asking employees to violate confidentiality, paying for insider access, relying on a single source.

The point is: channel checks are mosaic theory in action--combining public data with non-material observations to form a thesis. The mosaic is legal; the stolen piece is not.

Triangulating Alt Data with Fundamentals

Never trade on alt data alone. You triangulate: does this signal align with financials, guidance, and industry data?

The Framework:

1. Alt data signal: Web traffic up 18% month-over-month
2. Fundamental check: Prior guidance implied 10-12% revenue growth. Is 18% traffic plausible?
3. Industry context: Competitors showing similar trends?
4. Historical calibration: How did traffic-to-revenue conversion play out in prior quarters?

The test: can you explain divergence? Traffic up but revenue flat? Maybe conversion dropped. App downloads surging but engagement flat? Maybe paid acquisition is spiking with low-quality users. Why this matters: divergence is often a data quality issue, not an alpha signal.

Limitations and Lag Issues

Coverage Gaps: E-commerce has strong alt-data coverage; B2B software has weak coverage (enterprise deals not in panels); financial services are limited (AUM and fees invisible).

Lag: Alt data is not real-time. Web traffic: 7-14 day lag. Credit card data: 2-4 week lag. The point is: by the time you see the signal, fast-moving hedge funds may have already traded.

Signal Decay: Research shows alt-data alpha decays quickly. Web traffic signals lost 50% of predictive power within 2 years of becoming widely available.²

Worked Example: Web Traffic to Revenue Estimate

You analyze RetailCo (hypothetical), an e-commerce company. Consensus expects $1.15 billion Q4 revenue (+9% YoY).

Step 1 -- Gather Traffic Data:

• October visits: 42.3 million
• November visits: 48.1 million (+13.7% MoM)
• December pace: 51.2 million

Q4 total: ~141 million visits vs. ~125 million prior year (+12.8% YoY traffic growth).

Step 2 -- Estimate Revenue: Using 3.2% conversion rate and $85 AOV from prior filings:

• 141M x 3.2% x $85 = ~$384M/month
• Q4 implied: $1.15-$1.20 billion (supports a 2-4% beat)

Step 3 -- Triangulate:

• Credit card panel shows +11% spend growth (aligns)
• Competitor traffic up 8% (RetailCo outperforming)

Step 4 -- Size Position: Traffic data has 15-20% error bars. You assign 60% probability to beat, 30% in-line, 10% miss. This supports modest overweight--not aggressive positioning.

The durable lesson: alt data gave you a directional signal, but error bars prevented over-commitment--which is exactly how it should work.

Implementation Checklist (Tiered by ROI)

Essential (high ROI):

• Verify legal sourcing: can you explain where this data came from publicly?
• Check coverage: is this company well-represented in the data source?
• Triangulate: does alt data align with fundamentals?
• Assign error bars: what is the historical accuracy?

High-Impact (calibration workflow):

• Backtest signal vs. results for 4-8 quarters before trading
• Track signal decay: is alpha shrinking?
• Combine multiple alt-data sources to reduce noise

Optional (for active traders):

• Subscribe to real-time feeds ($10,000-$100,000+ annually)
• Build conversion/AOV models by company

The Durable Lesson

The durable lesson: alternative data and channel checks are tools for sharpening estimates, not replacing fundamental analysis. Practitioners who use alt data profitably treat it as one input among many--triangulating across sources and respecting the 30-50% noise rate that separates signal from speculation.

---

Footnotes

1. Alternativedata.org industry surveys and Greenwich Associates, Alternative Data in Institutional Investing (2021). ↩

2. McLean, R.D., & Pontiff, J. (2016). Does Academic Research Destroy Stock Return Predictability? Journal of Finance, 71(1), 5-32. ↩