I led Spotify’s data strategy under Daniel Ek at a time when music streaming was barely mainstream, and we hadn’t launched in America yet. My job was to move away from static Excel spreadsheets and toward a more dynamic way of tracking data for the first time – a far cry from the AI-powered analytics we see today, but back then we were inventing the wheel from first principles. 

The wheel was called Hadoop, software with a yellow stuffed elephant logo that haunts data analysts to this day. If you had a question about your data, you wrote a query, then waited hours for the system to tell you whether your query made sense in the first place. 

For the first time, we could track where users clicked, where they lingered, and use it to make better decisions. Learning how to do that was hard – it took us years to integrate this sudden influx of raw data into decision-making – but before we knew it, the platform surpassed 20 million subscribers with a level of precision that guesswork alone could never have gotten us to.

When I carried that muscle into venture capital half a decade later, it was clear that data, or an interest in it, had barely touched the investing world.

VCs don’t care about data

It was right as Stockholm’s tech scene was having its moment, post-Klarna and Spotify, hot on Silicon Valley’s heels in unicorns per capita. I was excited about what I could bring to the future of technology. At dinners, someone would mention a founder; nods were followed by anecdotes of a charming visionary. Those were the conversations that generated ‘hype’, which was often followed by increasingly large funding rounds. It didn’t matter about the company’s figures; it was FOMO on a once-in-a-generation opportunity. Some of these are still around today – many aren’t. 

Investing is, in many ways, an art. Being able to look at someone who has no idea what’s coming and know, before they do, that they’re about to reroute an industry. But that instinct can also be a flaw. Investors’ convictions get reinforced every time they are right, but written off without comment when they’re wrong, burying the fact that nearly two-thirds of venture deals never even return the original investment.

It’s all about the culture

Accusing investors of not tracking the misses may be uncharitable. Most firms keep a running post-mortem, but it is stored somewhere nobody typically looks, dusted off only when the next fund’s LP needs a slide on lessons learned.

Bessemer keeps a public anti-portfolio, showcasing famous deals they passed on – like Apple’s pre-IPO stock, which one partner dismissed at the time as “outrageously expensive” – but few other VCs are as self-reflective, particularly about the bets that failed. We use data to justify decisions after the fact, rarely before them. But without reflection, the next bet is exactly as blind as the last one. 

The reason for this is human. Data teams sit outside the main team, and don’t have the standing to change how a deal actually goes. Data’s role is to advise, and not decide; analysts fight for budget and wait to be consulted, while decisions happen in a room they’re not in.

… and the infrastructure 

The habit I’d built at Spotify, of making sure intuition was always backed by evidence, is what led me to build a tool to close this gap at my first VC. We called it Motherbrain, after the Nintendo character: it was the central intelligence at the heart of our investment decisions. In a pre-LLM world, it used signals like traction data, usage numbers or a spike in web traffic to predict opportunities before other investors even knew to look.

Within a few years, colleagues put a figure on what that meant: over $100 million of the fund’s investments had Motherbrain’s fingerprints on them. We tried to get portfolio companies to track performance better as well, and immediately realised that nearly none of them had the infrastructure to capture it in the first place.

Whose side is data on?

There is a tension at the core of venture capital. Entrepreneurs need funding, investors have it, and are eager to invest. On paper, that makes them the same team; in practice, each sees only half the picture. The house of cards is propped up by the founder’s account of how things are going, and the investor’s faith that they backed the right horse.

With better data, both sides can look at the same reality rather than two versions of it. So how do we do that? As firms race to adopt AI, startups that couldn’t track their own signals are bolting LLMs onto infrastructure never built to feed one. VCs do the same – it’s like pouring cooking oil in a gasoline car.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.