Chapter 3: The Data Edge

If you take one idea from this book into every investment decision you ever make, let it be this: in modern markets, the model is rarely the moat. The data is.

A brilliant neural network trained on public price data is a commodity. A mediocre model trained on data nobody else has, satellite photos of oil tankers, anonymized credit-card receipts, real-time foot traffic, is a license to print money, at least until everyone else buys the same feed. The entire alternative-data industry exists because of that sentence.

What "alternative data" actually is

"Alternative data" means anything that isn't the traditional trio of prices, financial statements, and economic releases. The categories that matter:

Geolocation and foot traffic. Firms like SafeGraph, Placer.ai, and Advan sell anonymized location data showing how many phones visited a store, a mall, a factory. If foot traffic at a retailer's stores is up 12% quarter over quarter while analysts expect flat sales, you know something the market doesn't, weeks before earnings.

Consumer transactions. Companies like Earnest Analytics, 1010data, and Second Measure aggregate anonymized credit- and debit-card spending. You can watch a company's revenue form in near-real time, transaction by transaction, while everyone else waits for the quarterly report.

Satellite imagery. Orbital Insight, RS Metrics, and SpaceKnow turn orbital photos into economic measurements. The classic case study: Orbital Insight estimates global oil inventories by measuring the shadows cast by floating-roof storage tanks. The roofs sink as tanks empty, and the shadows tell you by how much. That's a leading indicator for oil prices, derived from shadows, weeks ahead of official inventory reports.

Web-scraped activity. Thinknum, LinkUp, and Revelio Labs scrape job postings, prices, and product listings. A surge in job postings for "warehouse associates" at a retailer in October says something about holiday expectations. A company quietly raising prices across its website says something about margins.

News and social sentiment. RavenPack, RepRisk, FactSet Truvalue, and Social Market Analytics convert the entire firehose of news and social media into structured sentiment scores, the pipeline described in Chapter 1, running at institutional scale and speed.

How big this has gotten

A 2025 industry survey by Lowenstein Sandler found that 90% of investment advisers at private fund managers now use alternative data, up from 67% just a year earlier and 62% the year before that. This isn't experimental anymore. It's table stakes.

The market for alternative data was estimated around $7.9 billion in 2024, with vendor projections suggesting massive growth toward 2033. Treat vendor projections with appropriate skepticism. But whether the exact numbers hold, the direction is undeniable: the funds are spending enormous sums to know things first.

The 2026 reality: the moat moved

The data vendors won't put this in their brochures. Buying the data is now only half the battle, arguably the smaller half. The bigger costs are compute and engineers: cleaning the data, structuring it, testing whether a signal is real or noise, and doing it continuously as relationships decay.

Dirty data is worse than no data. A model trained on mislabeled transactions or biased samples doesn't just underperform. It hallucinates patterns with total confidence and loses real money acting on them. The funds that win are the ones with the data-engineering discipline to know when their data is lying to them.

This is also why the moat shifted. Ten years ago, simply having satellite imagery was an edge. Now everyone can buy satellite imagery. The edge is exclusivity, buying a feed nobody else gets, plus the infrastructure to exploit it before it decays. Speed of insight, not just possession of data.

What this means for you

I'll be direct: you will never have this data. Not the exclusive feeds, not the engineering teams, not the compute. A retail trader buying the same credit-card panel a fund buys is getting the commoditized version, months behind the funds that negotiated exclusivity, and paying a subscription price that only makes sense at institutional scale.

But the data edge still teaches something valuable. It explains why the market is so hard to beat. When you buy a stock because you liked the earnings call, you are trading against systems that watched the company's parking lots fill up, tracked its customers' credit cards, and read every word of that call through a sentiment model, weeks ago. That's not a reason to despair. It's a reason to be humble about stock-picking, to respect diversification, and to be deeply suspicious of anyone selling you "the same AI the big funds use." The funds' AI without the funds' data is a sports car without fuel.

There is one more edge the institutions have, and it's the one retail marketing lies about most: speed. That's the next chapter.