Our own numbers
Our median valuation rested on a handful of deals
For seven issues, our monthly report on venture funding printed the median valuation of the month’s rounds. When we rebuilt its history on a new data source, that one figure came back far from what we had published. On the rebuilt data, each month’s figure rested on 3 to 13 deals.
Deals behind each month
3 to 13
rounds with a disclosed valuation in the rebuilt data, out of 52 to 91 a month
November 2025, published
$500m
median valuation, as our report printed it
November 2025, rebuilt
$11,000m
the same figure on the new data, from 7 rounds
A figure in every issue
Our monthly report on venture funding rounds tracks how much was raised, in how many rounds, and by whom. From September 2025 to March 2026, seven issues also printed the median valuation: the middle value of the company valuations disclosed in that month’s rounds. It sat in the summary table beside total capital and median round size, and it had a chart of its own.
What looked wrong
Looking back, the figure moved in ways the rest of the report did not. It printed $290m in September 2025, $22.6m in October and $500m in November. The median round size over the same three months was $7.0m, $8.0m and $7.3m.
Our November issue read the jump as a sign that the month's funding had become more top-heavy. That reading did not hold up.
The check: how many deals sit behind each number
Our data provider stopped updating, and in September 2026 we moved the report to a public dataset of venture funding rounds. To keep the history comparable, we rebuilt every past month on the new data and set it beside what we had published.
The median round size came back close: within 7% of the published figure in six of the seven months, and within 18% in the seventh. The median valuation came back anywhere from a sixth of the published figure to 22 times it. November 2025 went from $500m to $11,000m.
So we counted the rounds behind it. Most funding rounds do not disclose a valuation. In the new data, across the seven months we had published, about one round in nine did: between 3 and 13 a month.
Funding rounds per month, September 2025 to March 2026
The same data as a table
| Month | All rounds | Disclosed a valuation | Share |
|---|---|---|---|
| Sep 2025 | 77 | 5 | 6.5% |
| Oct 2025 | 91 | 13 | 14.3% |
| Nov 2025 | 65 | 7 | 10.8% |
| Dec 2025 | 71 | 12 | 16.9% |
| Jan 2026 | 64 | 11 | 17.2% |
| Feb 2026 | 52 | 3 | 5.8% |
| Mar 2026 | 68 | 6 | 8.8% |
| Total | 488 | 57 | 11.7% |
In November 2025, seven rounds disclosed one. With seven numbers, the median is the fourth of them: one deal’s valuation. Which deals disclose a valuation changes from month to month, and the median changed with them. The earlier source carried valuations for only a small share of rounds too. Private companies mostly keep their valuation to themselves, so no choice of source would have fixed it.
What changed
We dropped the metric on 7 September 2026, and its chart in the report now shows the median round size. The size of a round is known for about four rounds in five, so each month’s median rests on dozens of deals, and it came back close to the published figure when we switched sources.
We also rebuilt the seven past issues on the new data, without the valuation figure, and each now carries a note saying why it went. The data behind the report still records how many valuations were disclosed each month, next to their median.
The general point
The same thing happens wherever a KPI is computed on a small group: average deal size for a sales team that closes six deals a month, or churn in a segment of 40 customers. The figure looks as solid as the ones beside it. The count beneath it is what says whether it can be read.
Start here
Is one of your KPIs resting on a handful of records?
Up to thirty minutes on one question, then a fixed proposal.