Small Markets, Outsize ROI

Almost every LP conversation we have had this year eventually arrives at the same question, usually asked politely: is my existing venture book just one massively correlated bet now?

It's a fair question. If you hold ten venture relationships and nine of them are underwriting enormous horizontal markets, winner-take-all dynamics, capital intensity by design, and an exit that depends on one of four acquirers or a generational IPO, you have diversified by manager. You have not diversified by underwriting. Ten managers making the same bet is one bet with ten sets of fees. A select few will back generational companies at the earliest stage. Most will not.

We think the more interesting exposure - right now - sits in the part of the market that the industry has been systematically trained to screen out.

TAM screening is backwards in our lane

Venture taught everyone to size the market/potential outcome first and everything else second. That instinct is right if you are deploying a multi-billion dollar platform fund, because you need a handful of outcomes large enough to move a very big number. And you have the checkwriting firepower to get meaningful ownership across more than a dozen option tickets per fund.

But it is worth being honest about what a large TAM actually tells you. It tells you how many well-capitalized people are going to show up to fight you for it. Big markets attract capital, capital attracts competitors, and competitors eventually compete on price. The math that gets quoted in the pitch, some small percentage of an enormous number, is usually the hardest math in all of technology to actually deliver.

Now run it the other way. A company that dominates a smaller, well-defined category - high gross margins, pricing power, almost no churn - is compounding real cash flow. A company holding a thin fraction of a massive category, against a field of well-capitalized competitors, is compounding a story that hasn't been proven out yet. One is a real business with real acquirers circling. The other is still betting the math plays out.

The test we actually run is customer ROI

The businesses that we help create displace a hard cost line rather than promise a soft productivity gain. A labor line. A loss line. They shrink vacancy, denied claims, compliance penalties, unbilled work. The budget already exists, which means the sale is a swap rather than a new line item, and the payback is measured in weeks rather than justified theoretically in a slide.

When the return is that lopsided, a lot of the things that quietly kill capital efficiency stop happening. Discounting is not part of the conversation. Renewal questions are not part of the conversation. Sales cycles compress because the buyer is doing arithmetic, not evaluating a vision. Churn, which is the single most underrated destroyer of venture outcomes, mostly disappears. We've been fortunate to back companies like the ones below where buyer math is so one-sided that it just disappears from consideration.

What that looks like in practice

EvolutionIQ, a Fund I investment, built AI-powered claims guidance for disability and workers' compensation carriers. Run the conventional screen and the market looks unappealing. The buyer universe is dozens of carriers, not thousands of companies, and it sits in a corner of insurance that Silicon Valley had shown no interest in. Meanwhile, many investments in "larger" parts of the consumer insurance landscape proved very disappointing. Run the ROI screen instead and it looks entirely different. Those carriers hold billions of dollars in claim reserves, and shortening claim duration while getting people back to work sooner is worth an enormous amount on a per-carrier basis. That is a product customers pay for, expand, and do not churn out of. CCC Intelligent Solutions acquired the company for $730 million, announced in the fourth quarter of 2024 (after Asymmetric invested in the first quarter of 2022). "Real" companies can have serious velocity. EIQ got taken out by a strategic buyer in the insurance software economy, not a hyperscaler, and did not require an IPO that the market had to cooperate on.

UpSmith deploys the same logic in an even less fashionable place. Wyatt Smith built software that helps skilled trades businesses recognize and reward the technician behavior that actually drives outcomes, integrated into the systems contractors already run. Ask how large the market is for paying HVAC technicians correctly and no venture committee gets excited. Ask a contractor what one retained senior technician is worth, or what a few points of close rate does to the year, and you get a number immediately. We invested at the earliest stage alongside a16z, in Dallas, in a category no platform fund is going to build a dedicated sales motion to attack.

We have a Fund II investment we will announce shortly that follows the same shape: a category most funds would size, wince at, and pass on, serving customers who can tell you to the dollar what the software is worth to them. That is the screen, not the market map.

Why this is newly investable

For most of the last two decades, a constrained TAM was a death sentence, and for a defensible reason. The fixed cost of building serious vertical software was high enough that a category with low billions of addressable spend could not carry a venture-scale outcome. The build ate the market.

AI collapsed that fixed cost. Categories that could not previously support the engineering now can, and they can support it with a fraction of the equity that the same business would have required in 2018.

The same smallness that used to disqualify these markets is what protects them. No mega-platform is going to build a dedicated go-to-market motion for a category this size. Distribution runs through people who have spent twenty years inside the industry and know who to call. The data compounds through repeated transactions, niche market brand authority and regulatory process, which is the variety that does not commoditize when the next model ships.

The exits look different, which is the point

These companies do not have one door. PE buyers want durable, high-margin, low-churn assets in unglamorous industries. Public and private strategics want the market share. Some of them can simply throw off cash while they wait.

That matters more to an LP than it does to a founder. Multiple exit paths change the timing and the probability of distributions, and distributions are the part of the return you actually get to spend.

What we are not claiming

This is not where the trillion dollar company gets built. If the frontier trade works the way consensus expects, this lane will not be the best performing thing you own, and we would not pretend otherwise.

The honest case is narrower and, we think, more useful: it does not require the same conditions to work. Entry price and ownership are the entire game here, which is also why fund size has to match the strategy. This approach is coherent at $137M. It is incoherent at $1B, and most of the capital in the market is now sized for the other side of the ledger.

So the question we would put back to LPs is this: if the AI capital cycle takes several years longer to convert into enterprise value than the current consensus assumes, what in your venture book still returns capital in the meantime?

If the honest answer is nothing, that is not a view. That is a concentration.

‍

Photo by Bertrand Borie on Unsplash

‍