Happy Thursday and welcome.

Today, we’ll go into frameworks and the type of analysis you can do, to get a better understanding of repeatability… or whether they're one deal in a trench coat.

Quick reminder to turn on “display images” in your email client or read it online.

A typical situation

A GP is in market for Fund V.

You review Fund III. Largely realised. 3.1x net. Top-quartile.

Then you pull deal-level data.

One investment - a software business - accounts for ~60% of total fund value. €35m in, ~€400m out. Exceptional outcome.

Remove it. Fund III is 1.7x net.

Still fine. But not stellar.

That’s the gap this piece is about.

The headline number doesn't tell you what you need to know

Fund-level TVPI and IRR compress everything into a single figure. That's useful for ranking. It's less useful for understanding.

What you actually need to know is: is this a repeatable process, or did the GP get lucky on one or two calls that papered over a mediocre portfolio?

Two things that help answer that question are the loss ratio and the return distribution test. Both are straightforward to run. Both require data that any GP should be able to provide. And together, they tell you more about the repeatability of the GPs returns than the headline numbers ever will.

1. The Loss Ratio

Start here because it's the simplest signal and the most honest.

Look at the GP's track record across all mature vintages. What share of investments ended in a loss or write-off - by count and by invested capital?

By count matters because it tells you how often the GP is wrong about a company. By capital matters because it tells you how much damage that costs when it happens.

You want both numbers, not just one. A GP who loses capital on 25% of deals by count but concentrates losses in very small positions looks different from one who took a large-sized bet that wiped out. The first might reflect a high-try, small-fail approach. The second tells you something about position sizing judgment. Or both might tell you something is off.

The ranges I use as reference points:

Under 10% by count and capital - the process is working and it's working consistently.

10–20% - acceptable. Losses are expected in private equity. The question is whether there's a pattern across them.

Above 20% - you're paying attention now. Above 30% - you want a compelling explanation before you go further.

But the number alone doesn't tell you enough. You need the second dimension.

2. The Symmetric Removal Test

This one I like because it's something you can build yourself in about twenty minutes with a deal-level data sheet.

Take the fund. Remove the top one or two deals by return contribution. Remove the bottom one or two deals by capital lost. Now look at what remains.

If the fund still holds up - still shows meaningful value creation across the rest of the portfolio - that's a GP who's building something. The exceptional deals are additive to a baseline that's already good. That's the sign of a process, not a lottery ticket.

If the fund collapses - falls to 1x or below, or drops below benchmark — the story changes. What you were looking at was concentrated risk that paid off, not compounding value creation across a portfolio.

The test is symmetric because you have to remove both tails. Some LPs only strip out the winners to check for home-run dependency. But stripping out the worst losses too gives you the baseline performance of the "middle" of the portfolio - the part that best represents what the GP actually does, repeatedly, on a normal deal.

What this isn't

I want to be honest about one thing: passing the balanced returns test doesn't mean the GP will repeat it.

A clean distribution across a portfolio of twelve deals is better than a home-run-dependent one. But it still describes one fund, in one market environment, with one set of entry conditions. The further back the vintage, the less it tells you about what this team will do in today's deals.

What it does tell you is whether the GP has been creating value consistently or has concentrated on the winners. That's a necessary input, not a sufficient one.

Headline numbers are easy to present. Return distribution is harder to dress up.

That’s why the line-by-line is so important to analyse.

At the start of April, FundFrame turned two. Instead of the classic “what we did” post, I’ll, instead focus on how things have changed.

Talk to customers more than you think you need to. This is startup advice #1, and I still underestimated it.

I came from 10+ years as an LP. I knew the problems. I'd lived them. At least, that’s what I told myself.

However, the specific friction, the exact workflow that breaks down: you can't model that from the outside. I tried. Doesn’t work.

So, if you’re starting out, don’t be like me. Get out there and talk to your peers.

The question you stop getting asked. Early on, every conversation had the same subtext: is FundFrame real, and will it still be here? At some point that question mostly disappeared. I didn't notice it happening. But one day, I noticed the absence of it.

I don’t fault anyone for asking it - I would have done the same.
But I don’t miss the question.

The product conviction gets stronger. Year one, I often had those doubtful days “what are you doing, Steffen”.

Year two, I don't. Not because everything went right. Because enough went right, in the right places, with the right people, that the doubt has less room. I know what we're building. I know it matters. That's not confidence for its own sake - it's just what happens when the work starts confirming the thesis.