How I'd Run My Portfolio If I Were New to Angel Investing

If you are an LP writing checks into VC SPVs, you are not running a fund. But you should treat the role the way a fund manager would, with similar discipline — specifically around portfolio construction.

You do not have a partner meeting. You do not have six weeks of diligence, a data room, customer calls, or a board seat. As an LP in SPVs, you typically get a memo, a lead, a cap table snippet, some growth numbers, a deck, and a close date. Then you decide in around a week if you want to invest.

Most new angels try to build conviction, size up the deals they like, save a reserve so they can follow, and end up with 8–15 names and no outlier. That is a fund strategy without fund information. As an LP in other people's SPVs, I would do something different. 

The constraint

A GP of a fund can underwrite the company. A new retail LP typically underwrites signals:

• Who is the VC leading

• Team pedigree

• Growth and traction relative to stage

• Social proof on the cap table and elsewhere

• Terms and valuation

• Whether you actually care about the space this year

And these signals are hard to judge.

Valuations get benchmarked against private market peers, which may be badly inflated, and against public comps, which can be just as inflated depending on where we are in the hype cycle.

And whether you care about the space is no easier. From what I've seen, very few LPs care about a category until it is already near hyperbolic. Like crypto, VC is reflexive. Being aware of that is how you decide whether this is the right strategy at all. Interest in a category is real, and it rotates: work-from-home trades and telehealth in 2020, SaaS in 2021 to Robotics/physical AI, defense, space, power today. You will have a sector bias. Just do not let this year's theme set the number of shots you take, or give you tunnel vision against the categories coming next.

VenCap has been backing top-tier VC funds since 1986, and they published the deal-level data on 11,350 companies across 259 funds — all of them raised by what they describe as some of the most successful firms in the industry. Over half those companies, 53%, came in below 1x cost, and nearly 23% were complete write-offs. Only 5.4% returned 10x or better, and just 1.1% returned the fund that backed them. Set that against the industry-wide benchmark: Correlation Ventures found 65% of deals returned less than the capital invested across 21,640 financings.

The point is not that tier-1 firms are bad at this. Their funds do fine if not great — in VenCap's own data, only 2% of core-manager funds returned less than 1x, against 22% across the industry. The point is that even the best firms in the world lose money on more than half their individual deals. 

How much of your net worth belongs here

Venture is a power-law sleeve, not a core holding. If I were starting over as a personal investor, I would cap it at low-to-mid single digits of liquid assets. Enough that an outlier changes your life. Not so much that a dead book changes your life the other way.

Fix the budget first. Then pick check size and construction.

A model for new SPV investors: access and frequency, not conviction and reserves

The angels who actually printed — Calacanis, Naval, FJ Labs, SV Angel — ran some version of the same machine. Many small, roughly equal first checks. Enough names that a power-law outcome is possible. Concentrate only after the company has produced new information.

Naval's version: see a huge funnel, pick a wide set that could be huge, keep the option to double down on the five that are. Calacanis's version: get 1 of ~100 right and ride it. AngelList looked at 10,000+ portfolios and found the median IRR rose with the number of investments. 

That is Model A. Model B is the concentrated, 1%-plus ownership, board-access super-angel. You are not Model B. You likely don’t have the information, or the time.

My most controversial take is that for LPs new to SPVs: the first check is the only check

I don't do this myself today. As a GP running my own vehicles and an angel who spends all day in the space, I've built confidence in specific areas, I see far more deals, I'm given far more information, and I understand the risks better than a new LP writing into other people's SPVs. 

As an individual LP new to the ecosystem, first-check-only is the better default. Three reasons:

Your outcome lives or dies on exposure to the outliers. Frequency is the variable you control. Ownership on name #7 that you sized up at seed is not.

Reserves sound disciplined and usually get spent badly. You save 40% "for winners," then miss the window, get squeezed on allocation, or follow a flat round because you already own it.

You cannot build real follow-on conviction. The next round arrives with a new deck, a new price, and the same information gap you had at seed. Re-underwriting without customers, a data room, or time is just a more expensive FOMO check into a company that you already have exposure to.

Let’s do a quick scenario. Same $300k: 15 investments × $20k versus 40 investments × $7.5k. Run both against a rough seed distribution — half go to zero, one in twenty returns 20x or better, one in a hundred returns 100x — and the two books have identical expected value. Everything that differs is distribution.

If both books catch a 100x, the concentrated one wins on dollars and it isn't close. That is the entire case for fewer, bigger checks, and it is a real case.

The problem is the "if." With 15 shots you catch a 100x about ~14% of the time. With 40 you catch one about a third of the time. You are trading a 2.7x better payoff for a far worse chance of having anything to collect it on, and the miss case is not a rounding error — it is the base case. Across the full distribution the 40-name book returns more than its capital 95% of the time versus 78%, and clears 3x more often.

Time diversity (Hamilton Lane's point, applied to SPVs)

Hamilton Lane's periodic table of returns is the clean version of this. It ranks pooled IRR by strategy for every vintage year, and venture does not sit still. It rotates from the top of the table to the bottom depending on the year you entered, and Hamilton Lane flags late-stage venture as one of the two most cyclical strategies they track. 

The year you showed up is doing a lot of the work, and if you showed up in a year like 2019 you probably did well and if you showed up in 2021 you probably did horribly. 

The number of checks buys company-level sampling. Time buys vintage sampling. You typically need both unless you catch the cycle at the perfect time. 

So in summary, how I would construct the book?

1. Budget first, then equal checks. Take the venture sleeve and deploy almost all of it as first checks. Example: a $300k sleeve becomes roughly $7.5k × 40 investments(or more).

2. Size every initial check the same. Do not upsize the obvious deal. The company that returns the book frequently doesn't look obvious, and obvious often just means overpriced. Equal checks are an admission that conviction without full information is a bad sizing input. You already cannot diligence the company.

3. Floor of 30–40+ names in a three-year cluster, not 8 high-conviction deals. A useful prior: a third go to zero, a handful return 1–5x, one or two pay for the rest. Below ~15–20 names, you are mostly asking whether the outlier happened to land in a tiny sample. ACA-style data shows larger books beating sub-10 books even after controlling for deal quality. 

4. Pace the vintage. 2021 SaaS felt like the only game in town at the time. It wasn't. Spreading entries across 2024, 2025, and 2026 is how you avoid a single-theme book tied to one valuation regime. If robotics is what you want to read about this year, fine — just don't let it become 90% of year one.

5. Follow-ons are not "support," and the default is no. This is probably my most controversial take, but for new investors it comes back to the same thing: you need exposure to the outlier, and the cheaper way to get it is more shots over more time.

The job you actually have

This all assumes picking well is very difficult, especially as a new investor. Almost nobody picks well when they are new, and it is compounded by the fact that you are not necessarily getting the access or information access or spending the time you would need to pick well anyway. Not off a lightweight memo and a close date.

We need to stop treating selection as the only job. You control how many checks you write, how big each one is, and over how long you spread them. Those three decisions are the most overlooked part of the job among the new LPs I see writing into SPVs, and they are the only part that is entirely yours.

Make them up front, before you start actively investing in privates.

✍️ Written by Zachary and Alex