# The Data Advantage: Will Proprietary Data Become the Most Valuable Corporate Asset?

Author: Norbert Biedrzycki
Published: 2021-12-15
Language: en
URL: https://norbertbiedrzycki.pl/en/the-data-advantage-will-proprietary-data-become-the-most-valuable/
Summary: Algorithms are increasingly available to everyone. Data is not. That simple asymmetry may reshape where the best AI investments are found.

**Algorithms are increasingly available to everyone. Data is not. That simple asymmetry may reshape where the best AI investments are found.**

For several years the AI conversation has been dominated by algorithms: which architecture, which framework, which research lab has published the latest breakthrough. Yet the more AI moves from research into business, the more obvious it becomes that the algorithm is rarely the scarcest ingredient.

Models are being open sourced. Cloud providers offer pre-trained services for vision, speech and language with a credit card. Talent is expensive but mobile. What cannot be downloaded is twenty years of claims data from an insurer, the transaction history of a bank, the maintenance records of a fleet of industrial machines or the purchasing behaviour of millions of customers of a retailer.

**Data as the real moat**

Several deals have already been made on this logic. In 2018 Roche paid around $1.9 billion for Flatiron Health, a company whose core asset was structured real world data from oncology practices. John Deere bought Blue River Technology, and this year it acquired Bear Flag Robotics, building on a uniquely rich base of data from millions of acres of farmland. These acquirers were not buying algorithms. They were buying the raw material for decades of algorithmic improvement.

For private equity this opens an interesting angle. Instead of competing for expensive AI startups, a fund could acquire established, often unfashionable businesses that sit on valuable datasets, then invest in the capabilities to monetise them. A mid sized logistics operator, a regional insurer or a specialised distributor may have exactly the kind of proprietary data that no startup can replicate.

**But data alone is not enough**

I want to be careful here, because this year has also given us a powerful counter-example.

In November, Zillow announced it was shutting down Zillow Offers, its algorithmic home buying business. Zillow arguably had one of the best real estate datasets in the United States and a well known pricing model. Yet when it used that model to buy homes at scale in a volatile market, it overpaid, accumulated inventory it could not sell profitably and announced write-downs of more than half a billion dollars along with significant layoffs.

The lesson is not that data is worthless. The lesson is that data creates value only in combination with three other things: a business model that tolerates the error rate of the model, an organisation that knows when not to trust the algorithm, and governance that monitors risk in real time. Zillow had the data. What failed was the link between prediction and capital at risk.

**How to assess a data advantage**

When I look at a company that claims its data is a strategic asset, I ask five questions. Is the data truly proprietary, or could a competitor buy something similar? Is it accumulated over time in a way that creates a flywheel, where more customers produce better data, which produces a better product, which attracts more customers? Is it legally usable for the intended purpose, given GDPR and sector specific regulation? Is it of sufficient quality, or will it take years of cleaning before it is useful? And does management have the operating capability to turn it into decisions?

In my experience as an operator, the last question is the one most often answered with optimism and least often with evidence.

**A shift in where AI value sits**

If this view is right, some of the most attractive AI investments of the coming decade may not look like AI companies at all. They will look like insurers, distributors, logistics operators and industrial service firms that discovered the value of what they already owned, and found the management talent to unlock it.

*Investment question: Could traditional businesses with proprietary data become more valuable AI investments than technology startups?*
