For coaches, mentors and incubator staff

---

The problem you already have

You sit down with a founder for an hour. Fifteen minutes goes on establishing what the company even is. Another fifteen on getting to a number they half-remember. You spend the remaining thirty giving advice that is necessarily generic, because you still do not know their COGS, their cap table, or how many prospects are actually on their list.

Two weeks later they come back — if they come back — and you re-establish the same context from scratch.

Meanwhile you give the same twenty pieces of advice to every founder you meet, in roughly the same order, because the same twenty things are missing every time.

startup-stack moves that first thirty minutes out of the session. The founder arrives with a stack, you read CONTEXT.md in four minutes before the call, and the hour is spent on judgment instead of archaeology.

The engagement model this assumes

Be explicit about it with founders on day one:

First session is free and deep. After that, continued time is earned by execution. Come back having done what we agreed and I will keep meeting you. Come back excited but empty-handed and this was a nice conversation.

This is not gatekeeping; it is triage. In most mentoring practices a large share of founders come once, learn something, and are never heard from again. Your hours are finite and they belong with the ones who execute. The weekly recap makes execution legible, so you are not making that judgment on vibes.

The recap is the mechanism. A founder who sends one every week — including the weeks that went badly — has told you everything you need to know about whether to keep investing your time.

The first session

Before: send them worksheets/startup-prep.md. Sixty minutes of their time, filled in before you meet, with best guesses where they have no evidence. It converts your discovery phase into their homework.

During: work through what they brought. Your job in session one is not to solve anything — it is to find which of the ten sections is emptiest and most dangerous. Usually it is 07-money (they do not know unit economics), 03-market (their competitive set is two names and both are wrong), or 05-gtm (their prospect list has five names on it).

After: the highest-leverage thing you can do is transcribe the session and run it through prompts/14-meeting-to-actions.md. It produces a summary, minutes, and owned action items with dates, and it writes the decisions back into the founder's stack.

This costs you ten minutes and does three things at once: it makes the advice durable rather than half-remembered, it creates a shared record of what was agreed, and it means the next session opens on "you committed to visiting 100 canteens — how many did you visit?" rather than "so, remind me where we got to."

Running a cohort

Standardise the stack, not the business. Same ten sections, same front matter, same recap shape across every company. It means you can read six recaps in fifteen minutes and see who is stuck.

Do not read master stacks. Ask for the derived copy — the founder runs the carve, restricted never leaves their machine. This matters legally and it matters for trust: a founder who thinks you can see their salary discussions will not write honestly about their salary discussions.

Run one session on the method itself. An hour, hands-on, everyone brings their raw files and bootstraps their stack in the room. Doing it together removes the single biggest barrier, which is not difficulty — it is that nobody starts.

Gate the next thing on the stack. If your programme offers grant capital, pilot introductions, lab access or demo-day slots, make a corrected stack and four consecutive weekly recaps the price of entry. This is not bureaucracy. A founder who cannot articulate their unit economics is not ready for the money, and giving it to them anyway is how grants get burned.

Watch section 3 across the cohort. "What did not move," aggregated over ten companies for four weeks, tells you exactly what your programme should be teaching next month. It is the best cohort-level signal you will get, and it is free.

If it is a portfolio rather than a cohort

Several companies, several coaches, and an institutional memory that has to survive any one of them leaving — that is a different problem from the one this page solves, and it needs a record the programme keeps for itself.

for-portfolios.md is the method: what to hold, why every session splits into a factual record and an attributed read, and how a maturity level routes a company to the right specialist. portfolio/ is the templates. exchange.md is the folder that connects a founder's stack to the programme's record without either side holding the other's master.

Everything on this page still applies. That layer only adds what your own team keeps.

The twenty questions

The stack is organised so that these get answered. They are the ones that come up in almost every session, and a founder who can answer all twenty from their own files is meaningfully ahead of one who cannot.

Company

1. What entity are you, and is it the right one for what you are trying to do next?
2. Who owns what, and does the cap table survive an investor looking at it?
3. Which advisors have you promised equity to, and is any of it in writing?

Customer

1. Describe your customer as a person, not a segment. When do they feel this problem?
2. What do they do today instead? What does that cost them?
3. How many of them have you actually spoken to?

Market

1. What is the market, how big is the bit you can serve, and where did those numbers come from?
2. Name five competitors. Now name the two you did not think of — the incumbent doing this badly, and the adjacent platform that could absorb you.
3. Why hasn't anyone else done this?

Product

1. What exists today that a customer could use, and what is still a slide?

Go to market

1. How many prospects are on your list? Five, or a hundred?
2. What happens when you personally go and knock on the door?
3. What are you spending on acquisition, and what does one customer cost you?

Operations

1. What is your capacity, and what breaks first if demand doubles?
2. Where does quality get checked, and by whom?

Money

1. What does one unit cost you, and what do you sell it for?
2. How much cash do you need to fulfil your biggest realistic order?
3. How many weeks of runway, and what is your plan B?

Capital

1. What are you raising for, at what milestone, and what does the investor get out in five years?
2. What non-dilutive money are you eligible for that you have not applied to?

If a founder cannot answer a question, that is a section of the stack to fill — not a lecture to give.

Advice that generalises

Patterns worth having ready, because they come up constantly:

"Your list has five names on it." Founders systematically under-scale outreach. Five prospects can easily yield zero. A hundred prospects, worked properly, reliably yields a handful. The instruction is not "improve your pitch," it is "make the list a hundred and come back."

"Go yourself." The founder personally visiting shops, societies, canteens, clinics or hospitals is worth more than any agency, at this stage, in every sector. It sharpens the pitch, it produces content, and it surfaces objections nobody would put in an email.

"Lock the paper before you allocate the equity." Advisors, marketing partners, technical friends — every one of them should have a one-page scope: role, deliverables, hours, consideration, vesting, term. Verbal equity promises are the most common unforced error in early companies. worksheets/advisor-scope.md.

"Documents before money." Founders ask for capital when what they lack is a costed process map. Give the grant after the documentation, not before — the documentation is what makes the grant productive.

"Phase zero has no overhead." Before machines, offices and hires, prove the economics through vendors and job work even at a worse margin. A 10% margin with no capex beats a 50% margin against a loan and an unproven customer.

"Step-test the spend." Not a 10× jump in one move. Double the daily budget, hold, measure — and repeat — long enough at each step to see whether cost per lead holds and whether operations can actually deliver the extra volume.

"Human in the loop." Any AI product that outputs a consequential recommendation — health, financial, legal — needs an explicit point where a qualified human validates before it reaches the end user. Founders building AI products almost always have to be told this once.

"Numbers must agree everywhere." Deck, website, plan and stack must carry the same headline figures. Inconsistency here is discovered by the one person in the room doing arithmetic, and it costs more credibility than the underlying number ever earns.

---

What to tell a founder who is resisting

They will say some version of: I don't have time to write documents, I need to be selling.

They are half right, and the answer is not to argue. It is:

Fine. Then do exactly two things. Fill in CONTEXT.md — one page, forty minutes. And send me a weekly recap. Nothing else.

Almost every founder who does those two things starts filling in the rest within a month, because the moment they need a deck, a trade sheet or a grant application, they discover it takes an hour instead of a week — and only because the context page exists.

Start there. The cathedral can wait.