Kevin was drowning. We had 150-200 full investments plus another 100-150 seed companies, and he was manually tracking performance metrics across all of them. Board decks were flooding in every month, each one formatted differently, with companies playing a shell game of which metrics they'd include or exclude. Meanwhile, partners were making investment decisions based on gut feel and scattered spreadsheets because no one had a clear view of how our portfolio was actually performing.

Quick Summary

We built a scalable portfolio monitoring system that automatically extracted and tracked performance metrics across 300+ companies, transforming how we measured and understood portfolio performance from manual chaos to automated intelligence.

The Problem

Board decks aren't GAAP accounting—they're whatever story a company wants to tell that month. One month they'd report revenue growth, the next month they'd switch to user metrics, and the month after that they'd focus on burn rate. We were trying to track performance across hundreds of companies with one person manually reading through decks, and it was impossible to get consistent, comparable data.

The real kicker? The most important updates weren't even in the board decks. They were coming through one-off emails, text messages, and random PDF updates sent between meetings. We were missing the actual performance story because we were only seeing the sanitized board presentation version.

My Approach

We mapped out the entire information flow and realized we had three core problems: data collection was manual and didn't scale, the source materials were inconsistent and incomplete, and we had no systematic way to capture the informal updates that mattered most.

My mental model was simple: if we could automate the extraction and create a repository that partners actually wanted to use, they'd start feeding us more and better data. But we needed to prove value first before asking people to change their workflows.

What I Built

I partnered with a third-party firm that used OCR + AI + human-in-the-loop processing to scrape board decks at scale. This freed Kevin from manual data entry to focus on the business logic—figuring out which metrics mattered, how to read between the lines of different company presentations, and doing high-level analysis.

But the real system was broader than just deck scraping. I built relationships with investing partners and portfolio support team members to capture those crucial one-off updates. We created a process where people would forward us emails, PDFs, even text messages with performance updates. Every week we'd run queries to find gaps in our data, then go hunt through emails and Google Drive folders to fill them in.

The end result was a queryable repository where we could generate reports, run lookbacks comparing board deck data versus informal updates, and actually benchmark companies against each other using consistent metrics over time.

Impact

For the first time, we could answer basic questions like "what percentage of our companies are hitting their revenue plans?" and "how much is our portfolio growing per quarter?" Partners started using the system because it gave them insights they couldn't get anywhere else. Kevin went from being buried in manual work to doing strategic analysis that informed investment decisions.

The system became self-reinforcing—the more value we provided through reports and insights, the more information people fed into it, which made our data even better.

Closing Thought

This taught me that the hardest part of portfolio monitoring isn't the technology—it's earning people's trust that you'll turn their messy, informal updates into something genuinely useful. Once you prove that, the data starts flowing on its own.