The rules of the earliest cheque are not the rules of venture.
Written for people who are excellent at something else and new to angel investing. It is the reasoning behind every dimension in a 710 Angels report — not advice on what to buy.
You are buying before there is anything to measure
That is the whole difference between angel and venture investing.
A venture fund at Series A can read a growth curve, a cohort table and a sales pipeline. At the angel stage those artefacts either do not exist or cover too short a period to mean much. You are underwriting a team, an insight and a plan — and you are doing it at a price low enough that being wrong most of the time is survivable.
A fund also has reserves, board seats and follow-on rights. You usually have none of those. Your protection is the entry price, the terms you accept, and how many bets you make.
The maths only works across a portfolio
Judged one company at a time, angel investing looks irrational. It is not meant to be judged that way.
Most early-stage companies return nothing. The returns of a whole portfolio typically come from a very small number of positions, which means the question in front of any single company is not "will this probably work" but "if it works, is it big enough to carry the ones that didn't".
Suppose you make ten equal investments. If seven return nothing, two return your money, and one returns fifteen times, the portfolio is up — driven entirely by one position.
Change that single company for a solid, profitable business that never multiplies in value, and the same portfolio loses money. This is why "good business" and "good angel investment" are not the same judgement, and why a ceiling on the valuation matters as much as the downside.
One practical consequence: a concentrated position in the very first company you like is the highest-risk way to start. Sizing is a decision you make before you fall for anything.
What diligence can and cannot establish here
You will not resolve the uncertainty. You are deciding which uncertainties you are willing to hold.
- Founders: relevant depth in this specific market, evidence of shipping, and how they behave when their own numbers are questioned.
- Market: whether the addressable slice is reasoned from the bottom up, or lifted from an industry report.
- Product evidence: what is actually in users' hands today versus what is described as coming.
- Capital efficiency: whether the use of funds adds up to the milestone claimed, and how many months of runway it really buys.
- Next-round raisability: whether the milestone this round funds is the one the next investor will require. If not, you are funding a bridge to nowhere.
- Clean paperwork: the cap table, outstanding SAFEs and their caps, IP assignment, and any founder departures.
The terms that actually change your outcome
Price gets the attention. These usually matter more.
- Valuation cap on a SAFE: the ceiling on the price your money converts at. Without one, a strong next round can convert you at a price that removes most of the benefit of being early.
- Discount: how much cheaper your conversion is than the next round's price. A discount without a cap is weak protection.
- Liquidation preference: who gets paid first if the company sells for a modest amount. At angel stage a plain 1× non-participating preference is the ordinary expectation.
- Pro rata rights: the right to invest again in later rounds so your stake is not diluted away just as the company starts working.
- Information rights: a regular update. Without it you may not learn a company is in trouble until it is over.
- Option pool timing: a pool created before the round dilutes the founders and the existing holders, not the incoming money. Check which side of the line it sits on.
Every term above has a one-line definition in the glossary.
The mistakes that repeat
Almost all of them are ways of skipping the portfolio question.
- Buying the demo. A polished product says the team can build; it says nothing about whether anyone will pay.
- Treating current revenue as the answer. Early revenue is a signal of pull, not a valuation basis, and a company can be profitable and still never multiply.
- Ignoring the next round. If nobody will price the next round higher, your paper gain never becomes a real one.
- Following a famous name without checking the terms — they may be in on a much better instrument than the one being offered to you.
- Deciding on a conversation. Founders are, by selection, persuasive. Write down what would have to be true, then check.
- No pacing. Spending an entire year's allocation in one month removes your ability to follow on with the company that works.
Two real cheques, re-read the way a report reads them
Famous early investments, reconstructed from what was knowable at the time — not from what happened later.
Thiel invested a reported $500,000 for roughly 10% of a college social network with no revenue, valuing it around $5 million. It became one of the best angel outcomes ever recorded.
Read it cold, the way a structured report would have had to:
- Evidence strength: strong on traction (documented, campus-by-campus usage), weak on monetisation (founder claim only). A report would have said exactly that — and the traction evidence was the part that mattered.
- What compounds: the network itself. Each new campus made the product harder to leave, a moat that widened with use. That, not the code, was the asset.
- Entry price: roughly $5 million for the category-defining network of its moment. The same company at $50 million would have been a far worse investment with the same evidence — price is part of the thesis, not a detail after it.
- Portfolio logic: this is the archetype of the position that carries a portfolio. It only "works" mathematically because the ceiling was unbounded and the entry was early.
Re-read in the AI era: one test has been added since — can AI flood the asset? Content is now free to generate, so a network whose value was user content alone would score lower today. Facebook's asset was real identity and real relationships, which are far harder to synthesize. The updated read asks of every network deal: which layer does AI commoditize — and is that the layer this company owns?
Before the Sequoia seed, the founders were introduced to a string of investors who passed. Some of the rejection emails survive: the market looked too small, strangers staying in strangers' homes looked unpalatable, the numbers looked modest.
Read the same material through the framework:
- The market-size mistake: readers priced "air mattresses in apartments" — the visible product — instead of the addressable slice of paid lodging. Bottom-up market reasoning is a diligence item precisely because top-down reads fail this way.
- The behavioural evidence was misread: cereal boxes and surviving a year unfunded read as desperation; they were documented proof of exactly the founder persistence diligence tries to establish.
- The falsifier that was available: talk to the New York hosts. One or two calls would have confirmed or killed the "real pull" claim — the fastest diligence there was, and almost nobody ran it.
- What it teaches: most passes were not wrong about risk; they were wrong about which evidence to weigh. A consistent rubric does not make you right — it makes your reasons explicit enough to check.
Re-read in the AI era: passing is easier to rationalize than ever — "an AI agent will intermediate all of this" is today's version of "the market is too small", and it stays untested until someone runs the falsifier. AI also sharpens the trap in the other direction: decks, projections and market maps are now cheap to produce beautifully, so the weight a report gives to documents over narrative is what separates a read from a vibe.
Figures are as publicly reported and rounded. The point is not that these outcomes were predictable — they were not. It is that the evidence available at the time was readable, and the discipline that reads it is the same one behind every dimension in a report here: weigh documents over claims, price the ceiling not the story, and write down what would prove you wrong.
- AI collapsed the cost of code, design and polished writing — three things that used to proxy for quality. Documented behaviour and real files are the remaining reliable signal, and the rubric weighs them accordingly.
- Add one question to every read: does AI strengthen or erode what this company owns? Infrastructure and accumulating assets tend to strengthen; thin features and content layers tend to erode.
- The rubric here carries these questions explicitly — AI replacement risk, time to copy in the AI era, what compounds — so the update is applied consistently to every deal, not only when you happen to remember to ask.
What the AI era added to the risk set
The traditional checklist no longer catches the fastest-moving failure mode.
- Displacement: could a general model, or a large AI company shipping an obvious feature, make this product unnecessary within a year or two.
- Time to copy: building software got dramatically cheaper. Any advantage that is only code is now a shorter advantage than it used to be.
- What compounds: proprietary data, distribution, workflow lock-in, regulatory clearance — something that widens the gap with use.
- Where the money goes: when most of a raise is engineering that a small team with modern tooling could now do far cheaper, the plan deserves a direct question.
- Valuation ceiling: whether the company can plausibly be worth several times more at the next round, or is priced today at what it will always be worth.
These are scored per sector and stage. A hardware company at pre-seed and a software company at seed are held to different rubrics, because the same question does not mean the same thing in both.
How to read a 710 Angels report
The labels around each judgement carry as much information as the number.
- Evidence strength: whether a judgement rests on a document, on partial proof, or only on the founder's assertion. Treat the last one as a question to ask, not a fact.
- Confidence: how much weight the whole read can carry, based on how complete the uploaded material is — not on how good the company is. Low confidence means ask for more, not decline.
- Not covered: the material never addressed the point, so it is excluded from the score rather than counted as a weakness. These are your first diligence questions.
- Falsifiers: the checks that would prove a judgement in the report wrong. Running one or two is usually the fastest diligence available.
- Founder responses: where founders have answered questions raised by the report, their replies appear alongside the original judgement.
The report is a structured read of the founder's own material. It does not recommend, rank or price anything, and a high score is not an endorsement. The decision stays with you.
After you invest, usefulness matters more than advice
The founder does not need another manager. They need an honest outside view and access they could not create alone.
- Ask what help is wanted before offering it. Your operating history is a source of hypotheses, not proof that the same answer fits this company.
- Keep updates light and consistent. Watch cash, the round milestone, the largest risk and the founder's specific requests; escalate only the signals that cannot wait.
- Keep your Portfolio Health diagnosis private. Convert it into a question, an evidence gap or an offer of support — never a score-based management order.
- Contribute where you have a real advantage: customer access, senior talent, sector pattern recognition, commercial partners, regulation or future capital.
- State disagreement clearly, propose a test, and respect the founder's operating decision. You still decide whether to follow on or commit more relationship capital.
The complete operating model — cadence, division of responsibility, introduction discipline, disagreement records, next-round preparation and copyable templates — is in the Founder × Investor partnership guide.
This page is educational material about how early-stage investing works. It is not investment, legal or tax advice, and nothing here is an offer or a recommendation to invest.
We do the reading. You make the decision.
We will never tell you a company is a buy. We produce a consistent, evidence-linked read of what the founder actually submitted, so that the time you spend is spent on companies inside your range and on questions that are still open.
We do not sell lists, and we do not use a low score to bury a company. Founders get the same report you do, and most of them use it to come back stronger.
Fewer meetings. Better ones.