The four criteria
- Demand: how much customers spend on this kind of business, whether that is growing or shrinking, and how much room is left for someone new.
- Margin: how much of each sale is left after the costs that come with it, such as product and card fees, and what owners report keeping.
- Start-up cost: what it costs to get going, such as equipment and getting it in place.
- Risk: the ways it most often goes wrong, starting with the failures owners mention most.
Each criterion gets one of four marks: Good, Mixed, Weak or Poor. The mark is drawn as a shape, so it reads without color: a filled circle for Good, a half-filled circle for Mixed, an empty circle for Weak, and a filled red circle with a bar across it for Poor. Each mark also carries its name for screen readers.
A score is a judgement from the evidence, written down with its reason. The table gives the line each mark is judged against, and says where vending got that mark.
| Criterion | Good | Mixed | Weak | Poor |
|---|---|---|---|---|
| Demand | Customers spend more each year, and there is room for someone new. | Customers spend a lot, but the market is shrinking or the good sites are taken (vending). | Spending is small or falling, and most of it goes to sellers already in place. | Few customers pay for this, and fewer each year. |
| Margin | Owners report keeping a clear share of each sale after product and fees. | Costs take much of each sale, but owners report a steady net. | Product and fees take a large share, and owners report a thin net (vending). | Most of each sale goes to costs before rent or your time. |
| Start-up cost | A few thousand dollars or less, and the equipment can be resold (vending). | More to start, or equipment that is hard to sell on. | A large outlay before the first sale, with little you can get back. | More than most first-time owners can raise, and little comes back if it fails. |
| Risk | Few common failures, and owners see them coming. | Known risks that most owners manage. | Common failures can cost you a site or what you paid, such as a lost location or an oversold route (vending). | Most who start lose money, or one common failure can end the business. |
On our page about whether vending machines are profitable, Margin is Weak because product costs about 35% to 56% of the price, card fees take 8% to 19% of a $1.50 card sale at 100 to 300 card sales a month, and the five owners who shared net figures keep a median of $136 per machine a month. Each figure links to its source.
The verdict
After the four scores, each page gives one verdict: Good, Only if or Hard. The page’s opening sentence gives it, and the score box at the top repeats it.
Good means the evidence supports the business for most people who can meet its start-up cost, with no condition that has to be met first. “Only if” means the business can work, but only under conditions the evidence points to, and the verdict names them. Hard means the evidence points against it for most people who try, even when some owners do well.
There is no formula that turns the four scores into a verdict. The verdict is a judgement from the scores, and each page says which scores drove it.
Vending machines are rated Only if: only if you get a location to say yes before you buy a machine, and you want a part-time restocking job rather than passive income. The vending page also says who it is not for: anyone who wants passive income, needs the money back within a few months, or can’t spend a few hours a week on it.
The page then says which scores drove it. For vending, a Good start-up cost and machines that resell limit the downside, while a Weak margin holds it back. Demand is Mixed and Risk is Weak, but no criterion scores Poor.
What we collect, and where from
We start with public data. For vending that meant the US Census Bureau, the Bureau of Labor Statistics, the FTC, the FDA and state and city rules and permit pages. We added trade body figures, a franchise disclosure, machine prices from sellers and listings, card reader fees and public commission contracts.
We then read owners who posted their own numbers: 13 public accounts for vending, 5 of them with a net figure.
We read every number from the page, PDF or data file it belongs to, never from a search snippet. Some sites don’t allow automated tools. When a site doesn’t, Tom Hill saves the pages himself and we work from his copies. For vending that meant Reddit threads.
How each number is dated and labelled
Every source shows the date we accessed it and, where it has one, its publication date. We check those publication dates. One survey of vending owners carries a 2024 date, but the Wayback Machine’s first copy is from October 2020, so we date it to about October 2020.
All figures are in US dollars. Anything in another currency is converted at the European Central Bank reference rate, taken from frankfurter.dev, and the source line says which day’s rate we used.
Where a figure is our own arithmetic, the working is kept beside it in the page’s source file, so the fact check can re-run it.
Sample sizes and their limits
Medians in our scores and tables carry their sample size. Owner accounts are self-reported, not audited, and mostly from people still in business, so they lean positive, and the page says so beside its owner figures. Used machine prices are asking prices, not sale prices.
What we never do
- We never invent a figure, a quote, a testimonial or a photo.
- We never name forum posters by their usernames. We describe them by role, such as “an ex-owner” or “a 7-year ex-operator”. People already named in the press or on their own channels keep their names.
- We never present receipts as income. Census receipts are sales before stock, commission, card fees and fuel, and the page says so.
- We never put Draper first unless it is the best option for that reader. On the vending page Draper’s free workshop sits alongside the free SBDC and SCORE help, not ahead of it, and the page lists what it can’t do.
How AI is used, and who checks it
Each rating page goes through the same steps, each done by a separate AI agent: a research brief, an evidence pack, an outline, a draft, a fact check, a plain-English edit and an independent review. Tom Hill then approves it. Until he does, the page stays out of search engines.
The fact check traces every figure to its evidence and re-opens sources live. On the vending page it checked about 155 figures, re-opened 20 sources and made 32 edits. In one case the draft said real income needs about 10 machines, but the ex-owner it quoted calls ten “the bare minimum” and says you need twenty or more for real income. The page now quotes him correctly.
The reviewer has not seen the page before. On the vending page it caught an answer saying owners keep “$70 to $136 a month at ordinary sites”, when the $70 is one ex-owner’s figure for his weakest sites.
Updates and corrections
There is no fixed schedule for re-checking figures. A page is re-checked when a source changes or someone reports an error, and the updated date in the byline changes when the page does. A corrected page goes live only after Tom approves the change. About and corrections has more.