Course progress
1
2
3
4
5
6
7
Lesson 3 of 7
3
Lesson 3 · 9 min · BriMindInvest Research Team

How Wrong Those Valuations Are — The Evidence

Graded against every covered company that has since listed: the median fund mark was materially below the first traded price, and the two worst cases were not errors at all but share-count changes. Why that distinction matters more than the error.

In this lesson you'll learn
What happens when a marked private company actually lists
Why 'error' is undefined when the share count changed
The median, the best case and the worst case, all shown
How to use a measured error band when sizing a position

The question almost nobody answers

Lesson 2 gave you a checkable private valuation. This lesson asks the follow-up that valuation providers rarely publish about themselves: when one of these companies finally listed, how close was the mark?

The only way to answer it is to wait for covered companies to go public and then compare. As of the latest build that has happened 6 times, at a median of 84 days between the last available mark and the first trade.

6
listings graded
3
with a comparable share count
3
recapitalised — scored separately
-38.2%
median error vs first close

6 is a small sample and no amount of presentation changes that. It is, however, the whole population — every covered company that has listed, with none removed for being embarrassing. Read it as a range of outcomes, not as a prediction.

Why two of these columns cannot be subtracted

A mark is dollars per share of a named instrument. A first-day close is dollars per share of newly listed common stock. Those are the same unit only if the share count did not change in between — and going public frequently changes it. Preferred converts, classes collapse into one, and a company will often split or reverse-split so the offer price lands in a conventional range.

When that happens the difference between the mark and the close is mostly the share-count change, not a valuation error. 3 of the 6 listings are in that state. They are kept in the table — hiding them would be worse — but flagged, and excluded from the headline statistic.

This is the detail that lets people publish enormous numbers about private-market gains. Compare a pre-split per-share mark with a post-split price and you can manufacture almost any return you like. Always ask whether the share count is the same on both sides.

The comparable listings

These are the 3 cases where the share count held and the subtraction is legitimate.

CompanyMarkOfferFirst closeMark errorDay-1 pop
CRCL
Circle Internet Group · 2025-06-05
$29.05$31.00$83.23-65.1%+168.5%
RDDT
Reddit · 2024-03-21
$31.15$34.00$50.44-38.2%+48.4%
CART
Instacart (Maplebear) · 2023-09-19
$32.50$30.00$33.70-3.6%+12.3%

A negative mark error means the funds valued the company below where it first closed. 3 of the graded events were marked below, 0 above, and 1 landed within 25% either way.

The recapitalised cases, shown but not scored

CompanyMarkFirst closeDay-1 popWhy excluded
SPCX$2,120.00$160.95+19.2%Share count changed before listing (≈10×) — per-share figures not comparable
FIG$26.39$115.50+250.0%Share count changed before listing — per-share figures not comparable
CRWV$939.85$40.000.0%Share count changed before listing — per-share figures not comparable

The two ends of the range

Best case
3.6%
Absolute error on the closest call. A handful of audited funds, weeks ahead of the listing, effectively priced the stock.
Worst case
2249.6%
Absolute error on the furthest miss — more than an order of magnitude, and a recapitalisation is most of it. We publish it because an error rate without its tail is marketing.

The median absolute error across the comparable set is 38.2%; across all events, including the recapitalised ones, it is 71.1%. The difference between those two numbers is the entire argument for separating the two groups.

What to actually do with this

The temptation is to turn the median error into a correction factor — "funds run 38% low, so mark everything up". Do not. The sample is tiny, the sign is not guaranteed, and the mechanism producing it is partly about how first-day prices are set rather than about value.

Two defensible uses:

  • As a confidence haircut. If your thesis needs the mark to be accurate within 20%, the measured error band says your thesis is not supported by this data.
  • As a warning about the first day. Look at the day-one pop column beside the mark error. A large pop and a large negative mark error are the same event seen twice — and it is the event you would have been buying into. That is Lesson 5.

The full table, including latest closes so you can see how these listings held up after the first day, is on the accuracy page.

Quick Knowledge Check
3 questions · test what you've just learned
1

A fund marked a company at $8.50 per share. It later listed and closed its first day at $31.00 — but the company did a 4-for-1 stock split on the way to market. What is the honest conclusion about the mark?

2

Across the graded listings, the median mark came in materially BELOW the first closing price. What is the most defensible reading?

3

Why is a sample of a handful of graded listings still worth publishing?

✓ Key takeaways from Lesson 3
Across the comparable listings the median mark sat -38.2% against the first closing price — consistently low, not randomly wrong.
A recapitalisation before listing makes per-share comparison meaningless. Those cases are scored separately rather than dropped or counted as errors.
The worst case in the sample is off by more than an order of magnitude. Any method that hides its worst case is not reporting an error rate.
Use the error band as a haircut on your own confidence, not as an adjustment factor to apply to marks.
The full accuracy table

Every graded listing with its mark, its filing date, its first close and its latest close — plus the ones we exclude and why.

See the scorecard →
← Lesson 2: Where a Private Company's Valuation Actually Comes FromNext: Lesson 4 — The Public Routes In: Proxies, Closed-End Funds & Discounts →