How to make a business plan for an early-stage startup (+ free pitch evaluator tool)
- BADideas.fund
- Jun 25
- 6 min read
Updated: Jul 6
The contrarian guide to how to make a business plan when you're pre-launch - validate the idea fast, kill it faster, and let one real buyer do what a 40-page document can't. From GTM operator and Nordigen co-founder Roberts Bernans.
I'm Jurģis - community & brand lead at BADideas.fund and host of the Bad Advice podcast. Every newsletter issue digs into multiple go-to-market problems from the show and hands you a free downloadable tool to fix it (a skill, an md spec, an AI VP of Sales or Marketing, etc.).
The short version, for the people in a hurry. If you're working out how to make a business plan for an early-stage startup, the most useful move is the opposite of what the templates tell you: don't write the long document first. At pre-launch your job isn't to plan, it's to learn whether the idea survives contact with one real buyer. Roberts Bernans - who co-founded Nordigen, scaled it across Europe and sold it to GoCardless - spent a full year building a startup he could have killed in an hour, because he validated late instead of early. The fix is a short loop: a couple of slides, one honest conversation with someone in the field, then kill or proceed. The plan is the output of validation, not the input. And before you pitch an investor, you can now see how we'd score your deck in minutes (free tool below).
I sat down with Roberts for an hour on our podcast, Bad Advice. Here's the go-to-market strategy framework that stuck.
The year-long business plan that should have been a one-hour kill
Before Nordigen, Roberts and his co-founder Rolands (who's now working on BirdyChat) had an idea: a marketplace where travellers could resell airline tickets they weren't going to use. The market logic looked fine - around 5% of people don't show up for their flights. So they did what founders are told to do. They built the business cards, the presentations, "all sorts of nonsense you don't need," and worked the idea for a year.
"We worked for a year on this idea, and we should have taken an hour to kill it"
The thing that finally killed it was a single conversation - they needed an airline as a launch partner, got in front of the AirBaltic board, and heard a flat no. A year of planning, one meeting to end it. The lesson Roberts pulled out is the one that should reshape how you think about how to make a business plan: the plan didn't protect them from anything. The market conversation they delayed for twelve months was the only document that mattered.
How to make a business plan that's actually a validation loop
Here's the reframe. A business plan for an early-stage startup isn't a document you write once and defend. It's a fast loop you run dozens of times until something sticks. When Roberts and Rolands reset after the airline flop, they built a weekly ritual: meet for a beer, each bring a list of ideas, pick one or two, and spend a week trying to validate them.
"Get a product as early as possible in front of your users, because only a user can tell you if it's good or bad"
The validation itself was deliberately cheap. No build. Maybe a couple of slides to illustrate the point.
"Don't build anything. Maybe build a couple of slides just to illustrate the point. Get it in front of a couple of people, validate, kill or proceed. Very simple."
They ran 100 ideas through that filter. Idea number 100-ish was Nordigen. That's the real shape of an early-stage business plan - not a tidy narrative you author in a vacuum, but the residue of a hundred fast kills.
Stop chasing statistical significance. One person is often enough.
The most common mistake Roberts sees in younger founders is over-engineering the validation step - the exact thing that makes a "proper" business plan feel productive while teaching you nothing.
"Young entrepreneurs over-engineer. Let's create a Google Form with 100 questions, build it out for weeks - then how do I find 100 people to get a statistically significant response? Sometimes it's enough with one person to tell you."
The point isn't to be rigorous for its own sake. It's to reduce ambiguity as fast and as cheaply as possible. One credible buyer who says "I'd never pay for this" has just saved you a year. He calls the mindset the cockroach: relentless, hard to kill, willing to ask.
"It doesn't cost you anything to invite someone on LinkedIn and say, 'Hey, I'm building a startup - can I grab you for a coffee chat?' Most will say no. Fine. Didn't cost you anything. But some will reply."
Whose advice belongs in the plan - and whose doesn't
A business plan absorbs advice from everywhere: accelerators, hackathons, investors, that one angel with strong opinions. Roberts' hard-won filter is to weigh the source before the advice. Early on, most people told him his idea was bad and Nordigen ranked at the bottom of pitch competitions. Following all of it would have been fatal.
"Most people who gave us advice had good intentions. It's just that their experience was either incomplete or too specific to be useful in our case."
His practical move: instead of asking an advisor "tell me about you," ask "tell me what you're great at," then take everything in context. The plan is yours to own - advisors inform it, they don't write it.
The AI mistake that wrecks a plan before it starts
Roberts now leads BADbrain, the AI infrastructure behind BADideas - effectively an AI interface to the fund that connects our 300+ operator and founder members to the help they need. So when he talks about AI in your business plan, it's worth listening.
"People start from the end. They go for the flashy things - autonomous agents - and don't realise nothing's working. You can't expect agents to do magic if you don't have any data, or it's structured the wrong way. The agent isn't working badly. What you're expecting from it is unfounded."
The plan-level takeaway: AI and automation are means to an end, not the strategy. "Every tool should have a purpose." Don't write "AI-powered" into your plan before you've written down what data you actually have.
The BADideas.fund pitch evaluator
Roberts' whole philosophy is that the best validation is one honest read from someone who actually decides. The trouble is that those people - investors - are exactly who you can't easily get a coffee with before you're ready. So we built the next best thing, and the same validate-fast logic runs underneath it.
Roberts matched a year of our real investment committee decisions - the scores and the written comments - against the pitch decks behind them, and extrapolated how BADideas actually makes decisions from a deck. Then he flipped that into a small tool: upload your pitch deck and get a glance at how BADideas would evaluate your business.
The feedback is personalised to how we score, so you're not getting boilerplate suggestions - you're getting a preview of how an early-stage B2B investor reads your business. Use it to find the holes before the real conversation.
Watch / listen to the full episode
Roberts goes deeper on the Nordigen journey, the lucky acquisition, hiring fast (and firing fast), and why he calls himself productively lazy.
Frequently asked questions
How do you make a business plan for an early-stage startup?
At the earliest stage, treat the business plan as a fast validation loop rather than a long document. Define the idea in a couple of slides, get it in front of one real potential buyer who can tell you it's good or bad, and then kill it or proceed. Run that loop across many ideas. The written plan is the output of validation, not the thing you write before validating.
Do I need a business plan to raise funding?
You need clarity far more than a long document. Investors back founders who can show who the buyer is, how acute the pain is, and evidence that real people want the thing. A clear, validated pitch deck does more than a 40-page plan. You can pre-check yours with the BADideas Pitch Evaluator before you pitch.
How much customer validation is enough before writing a plan?
Often far less than founders think. As Roberts Bernans puts it, sometimes one credible person in the field is enough to tell you whether an idea is worth pursuing. Chasing statistical significance with a 100-question survey usually delays the only signal that matters - a real buyer's honest reaction.
What's the biggest mistake founders make with AI in their plans?
Starting from the end. Founders reach for flashy autonomous agents before they have the data or data infrastructure to support them. AI should be a means to an end with a clear purpose, not the strategy itself written into the plan.
Work with BADideas.fund
We're an early-stage B2B fund across CEE and the Nordics, and everyone here has built companies before. We back founders whose drive borders on unreasonable and work alongside them on the GTM drift that kills startups between rounds - while there's still room to correct.
If you're raising and want operators who've walked your exact walk in the room with you:
