It's report season!

Last week, most pre-IPO secondary platforms posted their H1 2026 results. And I read them:

Caplight

Forge

Augment

EquityZen

The first thing that drew my attention: why are these reports so contradictory?

We platforms can't quite agree on almost anything:

• Performance of top-tier companies, as expressed in our indexes
• Concentration of the secondary market
• Secondary market discounts
• And most strikingly, the secondary price for the same company is different on every platform

Is it a bug or a feature? And what does the dispersion itself tell us about our market?

First, let's look at the data.

Concentration

• Augment: “a year ago, the six largest pre-IPO names on their platform drove close to half of all trading activity. This quarter they drove about a third — 29.9%, down from 47.4% in Q1.”
• LBXpro: LBX25 activity share is 56%.
• Hiive(cited in PitchBook's Q1 2026 report): the top 20 names account for 81% of trading value.

With concentration, the story might be fairly straightforward. If a platform mostly serves institutional investors (direct blocks, $1M+ checks), it sees a more concentrated market: institutions get to buy the best of what private markets offer. And a platform serving family offices and HNWIs (who mostly transact through SPVs) has to widen its universe of observable companies, because some of the most sought-after names never make it into an SPV at all — or never make it in on reasonable terms.

Secondary market median premium/discount

• Forge 0%
• EquityZen -38%
• LBXpro -40%

This is where it gets confusing. The off-the-shelf explanation: what if the median discount is also a proxy for the segment a platform serves? Institutional platforms see the best companies, and the best companies sell without a discount. And platforms handling smaller tickets see less popular companies with deeper discounts.

Clean theory. So I went to test it against company-level prices. If the hypothesis held, “institutional” platforms would consistently show better prices for the most popular names than “a more HNWI-oriented” platform.

Guess what?

No platform is systematically expensive or cheap.

Liquidity halves the dispersion but doesn't remove it. For recognized mega-caps, the median spread is 20.7%; for everything else, 46.4%. The tightest names are genuinely tight — OpenAI 3.7%, Databricks 5.7%, Shield AI 7.1%, Perplexity 8.9% — but Cohere (107%), Neuralink (96%), Mistral (73%) and Notion (68%) show that brand recognition is no guarantee of price agreement.

So my tidy "it's just segmentation" theory didn't survive contact with the data.

So what is going on?

You could explain the dispersion through differences in methodology, or even the business models behind each platform. But the simplest version is the most probable: each platform's "price" is really a measurement of its own order flow, and those flows barely overlap. Different methodologies are downstream of that — each platform is honestly describing a different slice of the market, not producing competing estimates of one true price.

The divergence isn't primarily a data-quality problem or a methodology problem. It's a structural feature of a pre-consensus market.

Will it be fixed?

The market almost certainly consolidates eventually. But the historical pattern — from corporate bonds to benchmark rates — is that consolidation comes from a mandated shared reference layer, not from one vendor's model quietly winning. No single platform is going to declare itself the price of record and have the rest of us fall in line.

Could the data be better (more useful) at the platform level in the meantime?

Well, platforms aren't really clarifying pricing for smaller-ticket audiences. The best any of us do for now is fold smaller tickets into the calculation — and even that only goes so far, because for marketplaces the real, all-in price of an SPV is often buried in fees and carry.

No platform can fully solve this alone. But a more transparent conversation about the SPV market would be a good start: which companies are even accessible in that universe, and what the real prices are for the part of it we can see clearly. For a market still searching for consensus, that's a more useful project than another index.