The time series is the asset.
Anyone can scrape EDGAR. Nobody can reproduce ten years of daily transaction history across every offering. Roughly 500,000 transaction-day records since launch day, May 16, 2016, collected as it happened, and impossible to rebuild after the fact. Across that record, 257 Reg CF issuers have gone on to raise $5.04B in institutional follow-on capital, roughly 6% of the dataset. Finding that 6% before the market does is the job.
Quantitative first. Human last.
The scoring framework
Every prospective investment runs through a multi-dimensional quantitative framework: financials, terms, traction, raise dynamics and market context. We name the discipline, not the weights. No company skips the score.
AI across the workflow
AI extracts structured data from each offering, normalizes terms, flags anomalies, and drafts the deal memo. It replaces hours of reading with minutes of review. It does not make the decision.
Human judgment closes the loop
An experienced investment committee reviews every candidate the system surfaces. Data proposes, people dispose.
A backtestable thesis
Because the data is historical, we can test a rule against what actually happened to thousands of past offerings. We measure what works instead of arguing for it.
Cast a wide net.
- 125-150early-stage companies per portfolio
- Breadthover concentration, in every round
- Neverleads; we follow documented rounds
- Valuationdiscipline at the core of every entry
- Follow-onoutcomes captured at scale
- LPsco-invest directly alongside the fund
Letter to investors.
The private capital markets are changing. Since Reg CF went live in 2016, thousands of companies have raised capital online every year, often at seed-stage valuations well below what traditional funds see, and with signals you cannot get anywhere else.
I have spent nearly a decade inside this market. As one of its pioneers, I began building what is now CCLEAR, the only complete record of U.S. investment crowdfunding. Today it tracks every offering since inception, with 125+ data points across roughly 10,000 companies.
That dataset is the foundation of D3VC. We run every prospective investment through a multi-dimensional quantitative scoring framework, apply AI across the workflow (from offering extraction to deal memo generation), and close the loop with experienced human investment committee review. Because the data is historical, the thesis is backtestable: we can measure what works rather than argue for it.
Our goal is not to pick unicorns. It is to construct a diversified portfolio of 125 to 150 early-stage companies with valuation discipline at the core, engineered to capture follow-on outcomes at scale. Because we take a partial position in each round and never lead, our LPs often co-invest directly alongside the fund, which makes D3VC as much a distribution channel for founders as it is a source of capital.
Sherwood NeissManaging Partner
Seven lessons from Fund I.
What the book taught us, written down so we keep doing it.
- 1
Read the clock before you read the number.
A position held at cost two months after the check tells you nothing; no one has had time to price it. The same position at cost two years later is information: the company hasn't raised, or hasn't raised up. Marks get more informative every quarter, so we read age first and value second.
- 2
A price nobody paid is not a price.
A founder can announce any valuation. What makes a number real is someone else's capital committed at it, documented. No document, no markup. That rule has kept our book honest when headlines were louder than filings.
- 3
The terms matter as much as the pick.
Most Reg CF instruments convert on a cap, often with a discount. Entering on good terms means a company can raise its next round at a lower headline valuation and our position still converts at or above what we paid. Selection gets the attention; terms do quiet work.
- 4
A valuation going up is not the same as your shares being worth more.
Headline valuations rise with new share issuance even when the price per share doesn't. We compare the price a new investor pays per share to the price we paid. That check has overturned gains that looked obvious from the press release.
- 5
Not every round goes up.
Not every round goes up. Some Fund I positions are carried below cost, and we mark them down the moment the evidence supports it. Pretending down rounds don't happen would mark the rest of the book wrong.
- 6
Silence is a signal.
Companies rarely announce bad news, but they do stop announcing good news. Missed annual reports, quieter investor updates, metrics that disappear from the deck. The gaps tell you as much as the numbers do.
- 7
Read the filing before the video.
The campaign video is produced to persuade. The Form C is a legal document the company can be held to. We start with the filing (financials, terms, use of proceeds, risk factors) and watch the video afterward, if at all.
And the conviction underneath all seven: cast a wide net.