How Much Money Do You Need to Start a Business? Real Starting Capital
There is no single figure, and any article that gives you one is guessing. Across the 406 businesses we index, 106 disclose a clean monthly revenue figure…
There is no single figure, and any article that gives you one is guessing. Across the 406 businesses we index, 106 disclose a clean monthly revenue figure spanning $6/mo to $2.2M/mo, median $27K/mo.
That $6/mo floor matters as much as the $2.2M/mo ceiling — both are real businesses in the same index. And 246 of the 406 are one-person operations, so the dominant starting capital here is one founder's time plus a small recurring stack, not a funded launch budget.
Table of contents
How much money do you really need to start?
Stop looking for an average. The labour line in most of these businesses is unpaid rather than budgeted.
Startup-cost articles quote agency estimates because agencies are the only party that publishes prices. Operators mostly do not. That asymmetry is why the internet's consensus number is high.
Our honest caveat, once: this index over-represents businesses that survived and then chose to talk about money. We record 38 documented failures across 406 businesses in 22 countries, nowhere near real attrition. Treat $27K/mo as a tool for comparing categories, never as a forecast.
What the real cost line items are
Four costs recur every month whatever you spend up front. Only platform fees and tooling have published prices; the founder's own hours silently dominate where 246 of 406 businesses are solo.
The SBA's startup cost worksheet is the right instrument precisely because it is dull: it separates one-time costs from recurring ones, the distinction every "it costs $X to start a business" headline collapses. Get platform fees from the operator's own published schedule, not a secondhand round number — fee tables are tiered and conditional. We break down an example in Etsy Shop Setup Fee.
| Cost line | Recurs? | Where the honest number comes from |
|---|---|---|
| Platform / marketplace fees | Per transaction | The platform's own published schedule, checked for your tier |
| Tooling and infrastructure | Monthly | Your own invoices — almost never disclosed by operators publicly |
| Paid acquisition | Monthly, and it grows with you | The line that separates $6/mo from $2.2M/mo |
| Founder time | Constantly, invisibly | Unpriced in 246 of the 406 indexed businesses |
Sizing a market rather than a budget? The U.S. Census quarterly e-commerce sales release is the primary source most "market size" figures trace back to.
What builders actually spent, by name
Revenue is well documented in our index. Spend is not. Each business below gets its figure, its evidence class, and a note where the cost base was never disclosed.
- ·[App Portfolio Studio Model](/projects/app-studio) — peak $2.2M/mo across 15 apps combined. Evidence class: 🗣 Founder-Reported. A marketing operator, a technical co-founder, and a developer community. Starting capital: not disclosed.
- ·[The Viral App Monetization Machine (100 Apps Dissected)](/projects/paywall-machine) — Cal AI and Lerna at $2M/mo each; LazyFit, CoinSnap and Impulse at $700K/mo each. Evidence class: ✅ Verified · Hard Data. Small teams throughout; the creator behind the analysis reports PuffCount at $1,300/day and Posted at $150K+/mo. Cost base: undisclosed throughout.
- ·[Retake AI Selfie Editor](/projects/retake-ai-selfie-editor) — ~$2M/mo, the creator citing third-party Sensor Tower data. Evidence class: 📎 Creator-Reported. Team and spend: not disclosed.
- ·[Gravl](/projects/gravl-ai-gym-workout-app) — $440K/mo with 70K+ subscribers. Evidence class: ✅ Verified · Hard Data. The one clear cost signal here: started with Julian and two others, now 13-14 people. Headcount followed revenue.
Cluely reports $500K/mo (~$6M ARR, founder-reported) with Roy Lee, co-founder Neil, ~13-14 people plus 60 contracted creators. And Leftclick reports ~$400K/month across businesses, founder-reported, with a 1 Second Copy content agency that peaked at $92K/month, Leftclick itself scaling to $72K in a month, and $100K+ in consulting. Browse all indexed ideas: earnings disclosed, spend withheld.
How much is each of these numbers actually worth?
Across the whole index: 57 third-party verified, 184 founder-reported, 121 creator-relayed, 44 unproven — 59% verified or founder-reported. Four in ten figures here come from someone who was never in the room.
| Evidence class | Count | What it means |
|---|---|---|
| ✅ Third-party verified | 57 | Someone outside the business saw hard data — a dashboard, an audited statement, an estimation platform |
| 🗣 Founder-reported | 184 | The operator stated the figure themselves, with their name attached |
| 📎 Creator-relayed | 121 | Repeated by a host or writer who did not personally verify it |
| 🔮 Unproven | 44 | A claim with no supporting artefact of any kind |
This is the argument the whole site rests on: a founder-reported figure and a verified figure are not the same claim, and stacking them in one list with identical formatting is a quiet form of dishonesty. Our AI Clone of Floe-Style Agency SaaS entry makes it concrete — the original benchmark does $1.3M/mo while the person demonstrating the clone does $0. Outrank is pushing toward $1M/mo (founder-reported, tiny team plus contractors in France); Quitter is at ~$500K/mo but secondhand via the Cluely interview, so creator-relayed, never to be cited with the weight of CoinSnap's verified $500K/mo. Rubric at how we grade the evidence.
The cost that decides the outcome is not the build
The build is the cheap part. The gap between the $6/mo floor and the $27K/mo median is almost never explained by how much someone spent shipping. It is explained by whether anyone found the thing.
Cleo is the cautionary version: a claimed $60K MRR in 53 days with zero proof, from a team whose earlier product Mentions did $20K MRR in month one — and the video making the claim is itself part of their launch funnel. Their stated lesson is that you cannot overspend on distribution. Take the lesson, discard the number.
Chai is the other side: $30M/yr, 21 people, a $450M valuation, and a cold start because the founder could not afford the GPT-3 bill and posted it online to crowdfund — 80K sign-ups in a day crashed the servers. Insufficient capital was the origin story, not the obstacle.
Rook did $47K over two months with a peak of $6,000+ in a single day, driven not by output volume but by an audience with a $24.5 RPM. Rate per view beat spend. For a category with an unusually transparent cost structure, see How To Start A Dropshipping Business.
How to use this without fooling yourself
- 1.Fill in the SBA worksheet before you fill in a business plan. With 246 of 406 indexed businesses running solo, your biggest recurring line is your own unpriced time — assign it an hourly rate and see whether the model still works.
- 2.Get platform fees from the platform. Do not budget off a percentage you read in an article. Check your actual tier this week.
- 3.Name your distribution channel before your build budget. Cleo's stated lesson was distribution spend; Rook's result came from a $24.5 RPM audience. If you cannot name the channel, the build number is meaningless.
- 4.Label every benchmark in your plan. If your projection leans on the $2M/mo Retake figure, write "creator-reported" next to it. If it leans on Gravl's $440K/mo, write "verified". Then use the SBA's guide to planning a business to build around the verified ones, and read more at the Revenue Reality hub.
Frequently asked questions
Is $20,000 enough to start a business?
Our data cannot confirm or deny a specific figure, because operators disclose revenue far more often than spend. What it does show: 246 of 406 indexed businesses are one-person operations, and the disclosing set starts as low as $6/mo in revenue. Capital is rarely the binding constraint in this index; distribution is.
Why do 90% of small businesses fail?
We will not repeat that statistic, because it is not in our data and we do not publish figures we cannot evidence. What we hold is 38 documented failures across 406 indexed businesses — and we say plainly that this understates reality, since the index over-represents businesses that survived and chose to disclose.
Is $10,000 enough to start a business?
Same answer as the $20,000 question, with the same evidence gap. The relevant signal is that the median disclosed revenue across 106 businesses is $27K/mo while the floor is $6/mo — the distribution of outcomes is enormous at any starting budget. Size your runway against the floor, not the median.
How reliable are these how much money do you need to start a business figures?
Partially. The whole index splits 57 third-party verified, 184 founder-reported, 121 creator-relayed, 44 unproven — 59% verified or founder-reported. Those classes apply to revenue. Cost data is weaker still: almost no operator in our 406 entries publishes starting capital, and we mark that as undisclosed rather than estimating it.
How many of the 406 indexed businesses actually disclose a monthly number?
106 of 406, across 22 countries, with a median of $27K/mo and a range of $6/mo to $2.2M/mo. The remaining 300 either never stated a figure or stated one too vague to record. That silence is data too — it is why single-number startup-cost averages deserve suspicion.