Paste a podcast
Back to the blog
Podex, Blog

how to validate a business idea: 6 tactics from andrew ng, pieter levels, alex hormozi and more

November 15, 202616 min read
“A lot of startups struggle not because we can't build whatever we want to build, but because we build something and it turns out nobody cares.”
Andrew Ng, Y Combinator

the short version

Try it on this episode

Why HubSpot’s Co-Founder Is Betting Against the SaaSpocalypse

a16z

or paste your own podcast

the six tactics

01

make the idea concrete enough to build this afternoon: Andrew Ng at Y Combinator

7 tactics in ~44 minutes (2637s). 437K views on YouTube, per the Podex page. Ng founded Coursera, ran Google Brain and now runs AI Fund. this play is tactic 01, Make every idea concrete enough to build this afternoon. on YouTube the talk is titled Andrew Ng: Building Faster with AI.

Ng says his studio only works on one kind of idea. on the captions: "a concrete product idea is one that's specified in enough detail that an engineer can go and build it." his bad example: "let's use AI to optimize healthcare assets you know that's actually not a concrete idea. It's too vague." his good one is software that lets patients book MRI machine slots online. he does not even claim that one is good. the point is that a concrete idea can be wrong, and you will notice fast: "having concrete ideas buys you speed." the Podex page sums up the trap: vague ideas get praise from friends because they are almost always right, so they never teach you anything.

the tactic: shrink the idea to one sentence a builder could act on today. who it is for, what it does, and how they get it.

try this week: write your idea as one sentence with a specific user, a specific job and a specific way they use it. if two friends would build two different things from that sentence, rewrite it until they would build the same thing.

Best forFounders who want ideas to leave the app.
Tactic 01

Andrew Ng, Y Combinator

Make every idea concrete enough to build this afternoon

The play

Rewrite your idea with enough detail that someone could start building it this afternoon.

Andrew Ng says

A concrete idea is specified in enough detail that an engineer can go build it, unlike a vague idea like using AI to optimize healthcare assets. A concrete example would be writing software to let hospitals let patients book MR machine slots online to optimize usage. That is concrete, something you could go build this afternoon. Vague ideas get kudos because if you tell your friends you should use AI to optimize the use of healthcare assets, everyone will say that's a great idea. When you're vague, you're almost always right, but when you're concrete, you may be right or wrong, either way is fine because you find out fast.

Try this

Rewrite your top product idea so specifically that an engineer could start building it this afternoon.

From Building Faster with AI ft. Andrew Ng →
02

run the yet test, then narrow the aperture: Dharmesh Shah on a16z

10 tactics in ~59 minutes (3546s). 21K views on YouTube, per the Podex page. Shah co-founded HubSpot and is its CTO. This play is tactic 01, Stay Off The Yellow Brick Road, plus tactic 04, Narrow The Aperture Before You Pick An Idea.

Shah's first check is about the reason you think you can win. if it is a gap in a big product, add one word. on the captions: "It's like, oh, ChatGPT can't do that yet. Does that ring true?" if it does, the gap may close and take your idea with it. his second check is size. founders pitch a huge market, and he answers: "tell me what you're going to do in the first 30 days." HubSpot itself started as a narrower test. their first plan was software for all small businesses, so they cut it: "Let's narrow the aperture and let's call it Legal Spot." talking to law firms killed that version. it was "not a fun market to sell into", and what the firms cared about was "the marketing and the website". that became HubSpot.

the tactic: test the reason you can win, then cut the market down to one you can reach this month. a narrow test teaches you faster than a big plan.

try this week: write one sentence on why your idea wins. add the word yet to it and read it out loud. then pick one narrow group you could talk to in 30 days (one industry, one city, one kind of user) and book three conversations with people in it.

Tactic 01

Dharmesh Shah, a16z

Stay Off The Yellow Brick Road

The play

Only build where a lab's own incentives, not just their current product gaps, keep them away, and recheck that assumption every quarter.

Dharmesh Shah says

I test an idea by appending the word yet to the competitive claim, and if it still rings true, the idea is a trap. I separate the product's current limits from what the company behind it could actually ship, and ambition is what keeps big labs out of certain markets, not inability.

Try this

Append the word yet to your competitive claim and reject the idea if it still rings true.

From Why HubSpot’s Co-Founder Is Betting Against the SaaSpocalypse →
03

write down what has to be true: Alex Hormozi with Jack Neel

12 tactics in ~109 minutes (6518s). 310K views on YouTube, per the Podex page. Hormozi co-founded Acquisition.com. This play is tactic 06, Test A Business Thesis With Need-To-Believes.

when Hormozi started Acquisition.com, he had no banking network and no proof the plan would work. on the captions: "I just thought of it in likelihoods." his thesis was that marketing could bring him deal flow: "So if I can just get more deal flow that wants to do deals with me, then I'll have a better pick". instead of judging the whole idea, he listed what had to be true. first, that he could make content that gets views. second, that views would turn into people who want to do deals with him: "So that was a second need to believe." he calls them "the two big ones". he also knew his downside: "worst case scenario, I'll just build an audience that doesn't generate deal flow and I'll figure something else out."

the tactic: break the idea into its need-to-believes. you cannot test an idea, but you can test two or three claims it depends on.

try this week: finish the sentence for this idea to work, I need to believe that... two or three times. next to each, write the cheapest test that could prove it wrong in a week, and the worst case if the whole idea fails. run the test for the riskiest one first.

Best forFounders who want ideas to leave the app.
04

get feedback in order of speed: Andrew Ng at Y Combinator

same Y Combinator talk. the Podex page lists this as tactic 05, Build the portfolio of feedback tactics, fastest first. Ng says AI has made building so fast that deciding what to build is now the slow part.

Ng's warning is about the real risk in most startups. on the captions: "we build something and it turns out nobody cares." so he runs a ladder of feedback, from fast and rough to slow and accurate. first: "the fastest tactic for getting feedback is look at the product yourself and just go by your gut." then "go ask three friends or teammates". then "ask three to 10 strangers". he learned to find them in hotel lobbies and coffee shops, where people working are often glad of a distraction, and says he has made product decisions there. after that come prototypes for 100 testers and A/B tests, which he calls "one of the slowest tactics in my menu".

the tactic: use the fastest test that can answer your question. save the slow, expensive tests for the questions the fast ones cannot answer.

try this week: show the one-sentence idea (or a rough mockup) to three friends, then to ten strangers in a coffee shop, a library or an online group where your users hang out. ask one question: would you use this, and what would you pay? write every answer down the same day.

05

ask for money, not an email: Pieter Levels on the Lex Fridman Podcast

6 tactics in ~224 minutes (13414s). 2.6M views on YouTube, per the Podex page. Levels has built over 40 products on his own. this play is tactic 02, Validate With A Payment Button, Not A Signup Form. the episode has a lot of swearing; we only quote clean lines.

Levels started his run of startups with a simple format. on the captions: "every month I would make something like a project and I would launch it with stripe so people could pay for it". a signup is not his test. he checks "if people actually sign up and not just sign up but if people actually pay money". and he is blunt about free plans for a solo founder: "free users generally don't convert". his advice is to "start asking people for money in the beginning", and not to price too low: "I would ask more than $10 per month". his summary for anyone starting out: "the important thing is that you validate that there's a market that there's a product that people want to pay for".

the tactic: replace the waitlist with a price. an email costs nothing to give. a card number tells you the problem is real.

try this week: make a one-page site with your one-sentence idea, a short demo or mockup, a real price and a payment link. if you cannot deliver yet, say when it ships and refund anyone who asks. send it to the ten strangers from entry 4 and count payments, not visits.

06

count only users who pay and keep using it: Alex Danilowicz and Jake Heller at Y Combinator

two YC episodes. Danilowicz (CEO of Magic Patterns): 12 tactics in ~48 minutes (2863s), 27K views, tactic 07, Validate With Ten Users Who Actually Pay And Use It. Heller (sold Casetext for $650M): 8 tactics in ~39 minutes (2365s), 262K views, tactic 07, Pilot revenue isn't real revenue.

Danilowicz spent months on pivots before Magic Patterns: "It took us about 9 months." the lesson he took, on the captions: "find 10 users who really love you." and he defines love strictly: "Where love is defined as they're paying you and they're using you." once the idea was right, "those 10 users came very quickly." Jake Heller adds the warning for bigger deals. on the captions, a cool demo can land a pilot, "But then it doesn't work in practice." his rule: "the sale does not end when they've written the check and definitely not when they started a pilot". as an angel investor he sees startups with big pilot numbers, and "a lot of those pilots are not converting to real revenue".

the tactic: set the bar at ten paying users who still use it. signups, pilots and praise come before that, not instead of it.

try this week: make a list of everyone who has paid or said yes so far. next to each name, write when they last used the product. if fewer than ten both paid and used it this month, your next job is to find out why, not to build more features.

a one-week plan to validate a business idea

pick by problem

Y CombinatorFREE

everyone says your idea sounds great, but nothing happens: Andrew Ng.

the play: rewrite the idea until someone could start building it this afternoon, so it can actually be tested.

the playWrite ten ideas, keep one, test it by Friday.
a16zFREE

your idea is a feature a big company does not have yet: Dharmesh Shah.

the play: add the word yet to your reason you win, then shrink the market to a group you can reach in 30 days.

the playWrite ten ideas, keep one, test it by Friday.
Jack NeelFREE

you cannot tell if the idea is good or bad: Alex Hormozi.

the play: list the two or three things that need to be true, then test the riskiest one first.

the playWrite ten ideas, keep one, test it by Friday.
Y CombinatorFREE

you are about to spend months building before anyone sees it: Andrew Ng.

the play: get feedback fastest first: your gut, three friends, ten strangers, then bigger tests.

the playWrite ten ideas, keep one, test it by Friday.
Lex FridmanFREE

you have a waitlist full of emails and no revenue: Pieter Levels.

the play: swap the signup form for a real price and a payment button.

the playWrite ten ideas, keep one, test it by Friday.
Y CombinatorFREE

you have pilots and signups, but you are not sure anyone needs it: Alex Danilowicz.

the play: count only the users who pay and use it this month. aim for ten.

the playWrite ten ideas, keep one, test it by Friday.

what this is not

want the locked tactics too?

every page above has more plays behind the free one: Ng's build twenty prototypes, Shah's charge real money on day one, Levels's A/B test on real users, Heller's price against replaced salaries.

The Tuesday Drop

one summary, every tuesday.

the tactics that matter, each with a step to run. unsubscribe anytime.

FAQ

How do I validate a business idea without spending money?

Start with the free tests Andrew Ng lists: your own gut, then three friends, then three to ten strangers in a coffee shop or online group. Then put up a one-page site with a payment link, like Pieter Levels does. A payment link and a free page builder cost little or nothing until someone pays.

How do I validate a business idea quickly?

Twenty shows, from My First Million to Huberman Lab.Make the idea concrete first. Andrew Ng says a vague idea cannot be tested fast, and "having concrete ideas buys you speed." Then use the fastest test that answers your question. Pieter Levels aims to launch a first version within about two weeks and check whether people pay.

Is a waitlist or a lot of signups enough to validate an idea?

Not on its own. Pieter Levels checks "if people actually sign up and not just sign up but if people actually pay money". Alex Danilowicz counts users who are paying and using the product. Signups are a reason to ask for money, not proof.

Should I charge from day one?

Pieter Levels and Dharmesh Shah both say yes. Shah says that with his newest product, "I monetized really early." Part of the reason was to learn pricing early, and he was surprised how many people put a card down. Levels says free users rarely convert for a solo founder.

How do I know if my business idea is good?

You find out by testing it, not by thinking longer. Kevin Ryan on The Tim Ferriss Show uses a gut filter first: "I start with a business crush, meaning there's an idea and I can't stop thinking about it." If it still holds after about two weeks, he moves. Then the tests above take over: need-to-believes, feedback, and payments.

Can I validate a business idea with AI?

AI makes building a prototype fast, but it cannot pay you. Andrew Ng says that with faster engineering, getting user feedback and deciding what to build "is increasingly the bottleneck". Use AI to build the demo or the page, then get the answer from real people with real money.