The Data Moat

In two prior posts, here and here I listed moats that I think are defensible in an AI-powered world. One more is truly proprietary data.

Why the “truly” qualifier? Because clever derivations from sources available to others are unlikely to be defensible. This would amount to a trade secret around the analysis, and trade secrets will become easier to rediscover. I believe the moat is limited to data that cannot be reproduced on the basis of sources available to others.

Businesses with network effects frequently have such proprietary data, but still deserve their own category. What types of proprietary data are not due to network effects?

A trading house that has captured some arbitrage opportunities may the signals that led to the opportunity from sight, taking them private, as it were. Sales histories may qualify, as may online tracking data, like location, site analytics, etc.

My guess as to the general rule is that sufficiently valuable data jointly produced from interactions between a product or service and users can remain private and defensible, assuming that the product or service provider retains the right to monetize the data.

So the new list reads:

• Brand
• Network Effects
• Physical Integration
• Sports
• Status Goods and Services
• Truly Proprietary Data