Edge in External Data

Most funds have a process. I’ve found few that are actively evolving that process to meet the full potential of AI.

For example, lots of data that informs a market is already sitting in the public internet, whether unstructured or embedded in filings, feeds, and documents. That data was typically something to reach for when a fund sits down to research a specific question. Today it can work for you continuously and proactively.

The premium on public data

A large share of what firms pay data vendors for is public information that someone else assembled first. The vendor’s moat is usually that it did the assembly and became the standard, not that it owns the underlying data. Modern tooling collapses the cost of that assembly and adds increasing value as you can customize its shape and accessibility.

Data that works for you

The next layer of value comes from automating the analysis on top of the data, across a corpus. Build the screens that actually matter, set statistical flags that surface what has moved, update the models the moment new data lands, encode your own view of the market so the system reasons the way you do, and have it reach out to you when something crosses a line.

What this looks like

FT Alphaville pulled apart Medallia’s deteriorating credit performance from public BDC filings, showing how lenders’ marks started to diverge and interest shifted from deferred PIK to cash payments. A revealing analysis, but performed on one borrower. Out of interest, I built the version that runs over the whole market.

Every quarter each business development company files a Schedule of Investments listing its private loans and the mark it holds them at.

• I turn every filing into a loan-level, cross-manager time series, resolve the same borrower across managers into one debt stack, and grade each manager’s marks against its peers and write-offs.
• From this data, we have an early-distress watchlist and an indicator of where marks may be soft, updated automatically as new filings are listed.
• Explore the full app here: bdc.credit

Any one-off analysis can be systematized into something that proactively watches, screens, and analyzes, and the same shape holds anywhere the data is available.