MeanCEO: Tech Startups and Startup Ideas

Minimum Viable Product Examples: What 11 Startups Built First, What Each Test Proved and What It Cost

TL;DR: Minimum viable product means the smallest thing you can put in front of real customers to learn whether your riskiest assumption holds. Dropbox used a video, Zappos used photos of shoes it did not own, and Deliveroo used three restaurants and the founder on a bike. Each tested ONE question, and the test type followed from it. Below: 11 sourced examples, seven test types compared, no-code tools with today's prices and a test card. The test picker gives you your own test type and pass line in a minute.
A minimum viable product is the cheapest, fastest version of an idea that real customers can react to with their time, email or money. In plain words: you build just enough to find out whether the most dangerous guess in your plan is true. That guess is your riskiest assumption, the belief that kills the business if wrong.
I have built companies in Europe for years and watched founders spend nine months on a first version that answered nobody's question. The usual listicles retell five stories and skip what each startup was trying to learn. This guide adds it. Every story and number was checked against its source on 8 November 2026.

What is a minimum viable product, and who coined the term?

Wikipedia credits the term to Frank Robinson, who coined it in 2001 at the consultancy SyncDev, which still frames it as the product that maximises return on risk for vendor and customer. Two people made it famous among startups:
  • Eric Ries wrote on his blog Startup Lessons Learned on 3 August 2009 that it lets a team "collect the maximum amount of validated learning about customers with the least effort."
  • Steve Blank called it the minimum feature set on 4 March 2010, a tactic to reduce wasted engineering hours, and warned founders against treating "minimum" as the goal.
The terms I use in this guide:
  • Riskiest assumption: the one belief that, if false, makes the rest of the plan pointless.
  • Early adopter: a customer who feels the problem so strongly that she tries an unfinished product.
  • Smoke test: a page or ad for a product that does not exist yet, measuring who signs up or tries to buy.
  • Concierge test: you deliver the result by hand to a few customers, openly, and learn what the job involves.
  • Wizard of Oz test: the customer sees a finished-looking product while a human does the work behind the curtain.
  • Pre-sale: you take payment before the product exists, with a delivery date and a refund promise.
My own definition is blunter: an experiment with a pass line written down before launch. Without that line you have a small product, and small products fail quietly.

How does the minimum viable product test picker work?

Pick the assumption that scares you most, your product type, budget, weeks and coding skills. The picker suggests a test type and a runner-up, names the startup from this article that tested the same way, lists what to build and skip, and writes a pass line you can edit, plus a test card to copy.
Minimum viable product test picker: which test, which pass line, what it costs

Start from the assumption that scares you most. The example below is a no-code software idea with €300 and four weeks. Change any field. Nothing leaves your browser.

Your test card (plain text)

How it works: the assumption and product type set the test type and runner-up; fewer weeks or less budget than the test needs moves you to the runner-up or a faster test, and "Neither" swaps builds for manual work. Pass lines are Violetta's rule of thumb, no study publishes one number for every market; the stop line is half the pass line. Cost and time ranges are estimates for a founder doing the work herself. Tool prices as shown on each pricing page, checked 8 November 2026: Carrd, Tally, Stripe (Dutch page), Softr, Glide, Bubble, Airtable, Zapier. Sources and date: TechCrunch (Eric Ries, 2011), Buffer blog (2011), Fortune (2012), Leigh Gallagher, "The Airbnb Story" (Penguin), Andrew Mason at Startup School (2014), The Next Web (2012, 2014), Guinness World Records, London Business School Review (2016), Raconteur (2024), TransferWise on VentureBeat (2017), Zalando history page, Sequoia Capital podcast with Markus Villig; all checked 8 November 2026.

The pass lines are my rules of thumb, since no study publishes one number for every market. Change them BEFORE the test runs.

What did 11 known startups build as their minimum viable product?

These minimum viable product examples come from founders' own accounts or reputable outlets, and four are European.

Dropbox: a demo video before the product worked for everyone

  • Built first: Drew Houston's three-minute screencast of the product as it was meant to work, full of in-jokes for the Digg tech crowd. Test type: explainer video.
  • Assumption tested: people want effortless file sync enough to sign up for something they cannot use yet.
  • What happened: Eric Ries reported in TechCrunch in 2011 that the video got more than 10,000 Diggs in 24 hours, and Houston said the beta waiting list went "from 5,000 people to 75,000 people literally overnight."
  • My lesson: it spoke to one community in its own language. A generic explainer proves nothing.

Buffer: two pages and a pricing click

  • Built first: Joel Gascoigne tweeted a two-page site, then slipped a pricing page between the two pages, so visitors picked a plan before leaving an email. Test type: landing page smoke test with a pricing step.
  • Assumption tested: people want scheduled social posts, and some will choose a paid plan.
  • What happened: Buffer's blog (16 February 2011) says he then built the product in seven weeks, launched on 30 November, had a paying customer within four days and saw around 4% upgrade to paid.
  • My lesson: the pricing click is the cheapest money signal there is. Add it to every landing page test.

Zappos: photos of shoes the founder did not own

  • Built first: Nick Swinmurn photographed the stock of Footwear Etc. in Sunnyvale and listed it online, promising the store, per Fortune, to buy any pair he sold at full price. Test type: Wizard of Oz.
  • Assumption tested: people will buy shoes online without trying them on.
  • What happened: "a few orders", then $150,000 from friends and colleagues and, later, $500,000 from Venture Frogs.
  • My lesson: he lost margin on every pair and bought the one answer that mattered.

Airbnb: three air mattresses for one conference

  • Built first: "AirBed & Breakfast", three air mattresses plus breakfast in the founders' San Francisco flat during a sold-out design conference in October 2007, according to Leigh Gallagher's "The Airbnb Story". Test type: single-feature build tied to one event.
  • Assumption tested: strangers will pay to sleep in a stranger's home.
  • What happened: three guests booked within a few days, at $80 a night per airbed.
  • My lesson: test when the problem is at its worst. Full hotels removed every excuse.

Groupon: WordPress, FileMaker and a mail script

  • Built first: Andrew Mason told Y Combinator's Startup School in 2014 that the team used a custom WordPress template, made vouchers in a FileMaker app and emailed them with a script that paused every 30 emails. Test type: piecemeal with existing tools.
  • Assumption tested: people will buy one local deal a day by email.
  • What happened: launch around October 2008 with two deals, sports-themed garter belts and two-for-one pizza. The pizza won, and Mason says they built it in a month.
  • My lesson: two offers on day one is a free A/B test.

Product Hunt: a 20-minute email list

  • Built first: Ryan Hoover invited a few dozen startup friends to add products to a daily digest on an existing tool, Linky Dink. He told The Next Web in 2014 it took 20 minutes to set up. Test type: piecemeal with existing tools.
  • Assumption tested: tech people want a daily list of new products picked by peers.
  • What happened: a few hundred people subscribed after he shared it on Twitter and Quibb. Only then came a website.
  • My lesson: the list tested a daily HABIT before any code existed, and habits are the hardest thing to fake.

Pebble: a pre-sale on Kickstarter

  • Built first: a Kickstarter campaign for an e-paper smartwatch, paid by backers before mass production. Test type: pre-sale.
  • Assumption tested: enough people will pay upfront to fund a first production run.
  • What happened: Guinness World Records lists $10,266,845 from 68,929 backers against a $100,000 goal, funded on 19 May 2012.
  • My lesson: for hardware, the pre-sale is test, funding and order book at once. Our guide to crowdfunding platforms in Europe compares the fees.

Deliveroo (UK): three restaurants and the founder on the bike

  • Built first: a London delivery service with three restaurants, founded in February 2013 by Will Shu and Greg Orlowski, per London Business School Review. Shu delivered orders himself. Test type: concierge.
  • Assumption tested: people will pay for delivery from good restaurants that do not deliver.
  • What happened: Shu told Raconteur in 2024, "I worked as a rider full-time for the first year of the business." His friends "kept ordering".
  • My lesson: the bike taught him what no survey reveals. A concierge test is research that pays you.

Wise (Estonia and UK): two friends swapping money

  • Built first: nothing. TransferWise's own sponsored story on VentureBeat says Taavet Hinrikus gave Kristo Käärmann euros and Käärmann paid pounds into Hinrikus's UK account at the Reuters mid-market rate, dodging bank costs of about 5%. Test type: concierge, founders as the first customers.
  • Assumption tested: people with opposite currency needs can swap at the real rate and both win.
  • What happened: TransferWise launched in January 2011, and The Next Web reported in April 2012 that it had saved early users over half a million pounds.
  • My lesson: they proved the economics with their own salaries. Be your own first customer when you can.

Zalando (Germany): an apartment, private phones and the post office

  • Built first: a shoe shop run from a shared Berlin apartment on Torstraße, autumn 2008. Zalando's history page says the founders' private mobile numbers were the hotline and they carried parcels to the post office themselves. Test type: piecemeal, manual operations behind a real shop.
  • Assumption tested: Germans will order shoes online. The company quotes the reaction it got: "Order shoes on the internet? That'll never work!"
  • What happened: no first-year numbers are published.
  • My lesson: answering the hotline yourself is the fastest research channel there is.

Bolt (Estonia): street pitches before the app existed

  • Built first: a pitch. Markus Villig told Sequoia Capital's podcast that at 19 he asked his parents for the first 5,000 euros to pay a developer for a prototype: "I didn't even have a product back then." Test type: concierge on the supply side, then a single-feature app.
  • Assumption tested: taxi drivers will join a new platform for extra income.
  • What happened: he pitched Tallinn drivers one by one in 2013. About 90% said no at first, and he ended with "a couple of dozen drivers lined up", enough to launch.
  • My lesson: in a marketplace, test the harder side first. A 90% rejection rate still passed.

Which test should you pick for your riskiest assumption?

Most founders start from what they can build. Start from what you need to learn instead. Let's break it down by assumption.
  1. Will people want it? Run a smoke test or an explainer video. Our guide on building a landing page that converts cold traffic covers the page, and building a waitlist that converts covers the emails.
  2. Will they pay? Run a pre-sale or add a pricing step, as Buffer did. Money is the only answer that cannot be polite. The playbook for pre-selling your product has the full method.
  3. Can we deliver it at a cost that works? Run a concierge or Wizard of Oz test and time every step.
  4. Will they keep using it? Build one feature or stitch existing tools together, and watch weeks 2 and 4. See our article on product market fit and the numbers that prove it.
Two filters on top: with fewer than two weeks, stay with a landing page or a pre-sale, and if nobody codes, swap the single-feature build for existing tools, as Groupon and Product Hunt did.

How do the minimum viable product test types compare?

Costs and times are my estimates for a founder doing the work herself; tool prices come from pricing pages checked on 8 November 2026.
Test typeWhat it testsCash cost (estimate)Time (estimate)What it cannot tell youStartup in this article
Smoke test / landing pageInterest in your message€0 to €100 plus ads1 to 2 weeksWhether anyone pays or staysBuffer
Explainer videoInterest in a product nobody can try yet€0 to €3001 to 3 weeksWhether the real product satisfiesDropbox
Concierge testWhether the result is worth paying for€0 plus your hours2 to 6 weeksWhether it works without your handsDeliveroo, Wise, Bolt
Wizard of Oz testUse of a finished-looking product€0 to €2002 to 4 weeksWhether automation matches the manual costZappos
Pre-saleWillingness to pay upfront€0 plus card fees1 to 3 weeksWhether buyers stayPebble
Single-feature buildRepeat use of one job€300 to €3,0004 to 8 weeksDemand outside your nicheAirbnb
Piecemeal with existing toolsThe whole service on rented tools€0 to €100 a month1 to 3 weeksMargins at volumeGroupon, Product Hunt, Zalando
On Stripe's Dutch pricing page, Payment Links cost nothing extra and a standard EEA card payment costs 1.5% + €0.25, so a €49 pre-order costs about €0.99 in fees. For a public pre-sale with a deadline, our crowdfunding tips and week-by-week campaign plan covers the goal, the fees and the pre-launch work.

How do you build a minimum viable product without code?

How to build a minimum viable product without a developer:
  1. Write the test card first (template below): one assumption, metric, pass line and stop rule.
  2. Pick the test type from the table above before you pick a tool.
  3. List what to skip. Logins, settings, admin panels and mobile apps wait. Our guide to the services-to-product pivot keeps cash coming in meanwhile.
  4. Assemble the smallest stack from the table below, or see our tool stack for building on a bootstrap budget.
  5. Recruit 20 to 50 early adopters by hand. Our notes on lean customer acquisition for bootstrapped startups say where.
  6. Run it, then decide: continue, change one thing, or stop.
ToolGood forCheapest paid plan, as shown on 8 Nov 2026
CarrdOne-page smoke testPro Standard $19 / year, first plan with forms
FramerPolished landing page, custom domain$10 per month (free plan: no custom domain)
TallySignup, survey and order formsFree within fair use; Pro €20 billed monthly
Stripe Payment LinksPre-sales and depositsNo plan fee; EEA cards 1.5% + €0.25 per payment
AirtableBack office for manual testsTeam $20 per user per month, billed annually
SoftrClient portal on top of your data$19 per month
GlideMobile-friendly app from a spreadsheet$25 per month
BubbleWeb app with logins and paymentsStarter $32 per month, billed annually (needed to go live)
ZapierGluing forms, sheets and paymentsPro from $19.99 per month billed annually, $29.99 monthly
My first-test stack costs under €30: Carrd or Tally, Stripe, a spreadsheet and my own hands. Our breakdown of startup costs in 2026 covers the rest of a launch.

What is the difference between a minimum viable product, a prototype and a proof of concept?

Minimum viable product vs prototype comes down to who touches it and what you measure.
Proof of conceptPrototypeMinimum viable product
Question it answersCan this be built at all?How will it look and work?Will customers want, pay for or keep using it?
Who sees itThe team, sometimes an investorTest users in a sessionReal customers in real conditions
What it isA technical experimentA clickable mock-up or modelA working offer, even if manual behind the scenes
Success looks likeIt works in the labPeople understand and complete the taskPeople sign up, pay or come back
A proof of concept reduces technical risk, a prototype design risk and a minimum viable product market risk, the one that kills most. CB Insights studied 431 VC-backed startups shut down since 2023 and reported in March 2026 that 43% of the 385 with known causes cited poor product-market fit.

What mistakes kill a minimum viable product test?

The mistakes I see in first tests:
  • No pass line before launch. Decide afterwards and you will decide in your own favour. Our guide to preventing confirmation bias when validating an idea has more traps.
  • Testing two assumptions at once. A fail cannot show which one broke.
  • Counting compliments. Count emails, booked calls and payments.
  • Building the easy side of a marketplace. Bolt went after drivers first because supply was the hard part.
  • Hiding the price. Without a price, every signup is a maybe.
  • Too few people. My minimum: 100 visitors on a smoke test, 10 customers on a concierge test.
  • Polishing before learning. Logos and names can wait; later, the business name generator gives you options in a minute.
  • Refusing to stop. Honour the stop rule. Our guide on how to pivot without burning your bootstrap runway covers what comes after a fail.

What should your minimum viable product test card say?

Copy it, fill in the brackets and share it before launch:
Minimum viable product test card
  • We believe that: [customer] will [want / pay for / keep using] [offer].
  • Riskiest assumption: [the one belief that kills the plan if false]
  • Test type: [smoke test / explainer video / concierge / Wizard of Oz / pre-sale / single feature / existing tools]
  • We build: [three things, no more]
  • We skip: [everything else, named]
  • We reach: [number] people through [channel] in [weeks] weeks
  • Metric: [signups / payments / repeat use] out of [visitors / people asked / users]
  • Pass line, set today: [number]%
  • Stop rule: below half the pass line, we stop or change ONE thing and rerun once.
  • Budget: €[amount]. Owner: [name]. Decision date: [date].
For interviews alongside the test, see our article on customer research for product development.
Start with a sharper idea. No idea worth testing yet? Try the business idea generator. Map customer, problem and revenue with the free lean canvas template, and see which grants can pay for your tests with the non-dilutive funding finder.

What should you do with your minimum viable product this week?

Every startup here tested one belief with the cheapest thing that could prove it false, from Dropbox's video to Pebble's pre-sale. None started with the full product.
Next steps:
  • Today: write your riskiest assumption in one sentence and run it through the test picker above.
  • This week: set the pass line, build the three things on your card and reach 20 people by hand.
  • In two to four weeks: compare the result with the pass line and decide.
My prediction for 2027: AI builders make code nearly free, so more founders will skip the test and build everything in a weekend. They will still fail because nobody wanted it. If you are still choosing what to build, our guide on validating a startup idea with less than €1,000 is the next read.

FAQ: minimum viable product

What is a minimum viable product in simple terms?

It is the smallest version of your idea that real customers can respond to with their email, time or money, built to test your riskiest assumption cheaply.

Who invented the minimum viable product?

Wikipedia credits Frank Robinson of SyncDev, in 2001. Eric Ries (2009) and Steve Blank (2010) made it famous among startups.

What are the most famous minimum viable product examples?

Dropbox's demo video, Zappos's borrowed shoe photos, Airbnb's air mattresses and Pebble's Kickstarter pre-sale. European examples include Deliveroo, Wise, Zalando and Bolt.

What is the difference between a minimum viable product and a prototype?

A prototype shows how something will look and work to test users. A minimum viable product goes to real customers and measures whether they sign up, pay or come back.

How much does a minimum viable product cost?

By my estimates, a smoke test or concierge test costs €0 to about €100 in cash, and a no-code single-feature app a few hundred euros (Bubble's Starter plan is $32 per month billed annually, checked 8 November 2026).

How long should a minimum viable product test run?

One to three weeks for a landing page or pre-sale, two to six for a concierge or Wizard of Oz test, four to eight for a single-feature build. Set the end date first.

What is a Wizard of Oz test?

Customers see a finished-looking product while a person does the work behind the scenes, as Zappos did by buying each ordered pair from a local store.

What is a concierge test?

You deliver the result by hand to a few customers who know a person does the work, as Deliveroo's founder did on his bike for the first year.

About the Author

Violetta Bonenkamp, also known as MeanCEO, is an experienced startup founder with an impressive educational background including an MBA and four other higher education degrees. She has over 20 years of work experience across multiple countries, including 5 years as a solopreneur and serial entrepreneur. Throughout her startup experience she has applied for multiple startup grants at the EU level, in the Netherlands and Malta, and her startups received quite a few of those. She’s been living, studying and working in many countries around the globe and her extensive multicultural experience has influenced her immensely.
2026-11-08 08:53 Guides