\"Passive\" income for freelancers: the unit economics nobody shows you

The pitch for passive income is that you build the thing once and it pays you afterward, so the ceiling on your earnings stops being the number of hours in your week. The pitch is not wrong about the mechanism. It is almost always wrong about the size of the first payment and the length of the wait.

Here is the wait, in the cheapest category available. A Kindle ebook priced at $9.99 earns the 70% royalty option, which pays 70% of list price less delivery costs of $0.15 per megabyte in the US store (Amazon KDP royalty terms). On a 1 MB file that is 0.7 × ($9.99 − $0.15) = $6.89 a copy. Suppose the book takes 60 hours to write and edit, and suppose you would otherwise have billed those hours at $80 — both figures are illustrations, so substitute your own. Those 60 hours cost you $4,800 of client revenue you did not invoice. At $6.89 a copy, the book repays that at copy number 697.

That is the arithmetic missing from almost every article on this subject, and it is the only arithmetic that decides anything.

The build hours are billable hours, and they are the real price

Money spent on a course platform is trivially small next to the cost that never appears on a statement: the hours. A freelancer's build hours are not spare capacity. They are billable hours redirected, which means every product carries a price tag equal to hours × your rate, paid up front, in full, before the first sale.

So the useful unit for all of this is billable hours forgone. Not months to break even, which flatters anything with a long horizon, and not revenue, which ignores that you paid for it with your own capacity. Hours forgone gives you a number you can compare directly against the alternative use of the same hours.

Two things make the comparison honest. First, count the whole build: not just making the thing but the sales page, the payment setup, the tax treatment of a new revenue line, and the three weekends of tinkering that follow launch. Most people undercount by half. Second, count upkeep separately, because it never stops. Support email, refunds, updates when the software your course teaches changes its interface, replacing a payment integration. Upkeep is the reason "passive" is the wrong word: the correct word is deferred, and the deferral has a maintenance bill attached. The cheapest way to know either number is to run the timer on product work exactly as you would on a client project, since time tracking that feeds invoices will also tell you what the build really cost when you are done arguing with yourself about it.

If you do not know what your hours are worth, the whole exercise collapses, because you have no exchange rate. An effective hourly rate check on your last three finished projects gives you the honest figure, which is usually lower than your quoted rate. Freelance rates: the number that matters isn't the one you quote explains why the gap exists.

Five categories, priced in hours rather than in dreams

Below, the same illustration runs through every row: $80 an hour, with build and upkeep hours chosen to be plausible rather than measured. The platform fees are real and linked, and each platform charges them worldwide, so read the dollar figures as your own currency — the ratios are what carries. The one exception is the Kindle price band, which is specific to Amazon.com. The last column is what it takes in year one to get back to zero, counting the forgone billing and the upkeep but not counting the value of your time as a rounding error, because it isn't one.

The course, worked all the way through

Courses attract the most attention and carry the worst ratio, so they deserve the arithmetic in full. On Teachable's Starter plan at $39 a month with a 7.5% transaction fee, a $149 course nets $149 − $11.18 = $137.83, and the plan costs $468 a year. Against 120 build hours ($9,600) and four upkeep hours a month ($3,840), year one needs ($9,600 + $3,840 + $468) ÷ $137.83 = 101 sales to get back to zero.

The plan choice is its own small decision with a clean threshold. Builder costs $89 a month instead of $39, an extra $600 a year, but charges 0% per transaction, saving $11.18 a sale. $600 ÷ $11.18 = 54, so above about 54 sales a year the more expensive plan is cheaper. That is the kind of arithmetic worth doing once and then forgetting.

Now the part that decides the outcome: where 101 buyers come from. Nobody publishes a defensible conversion benchmark for freelancer-audience course sales, and any figure you find in a blog post is a marketing artifact, so treat the number below as a placeholder for a number you will have to measure. At a 1% conversion from qualified viewer to buyer, 101 sales means putting the offer in front of roughly 10,100 people in a year. At 2%, roughly 5,050. If you cannot name today where those people come from, the course is not a revenue plan, it is a hobby with a landing page.

The template pack, which is the honest small bet

At $39 with Gumroad's direct-link fee of 10% + $0.50, you keep $34.60. Against 25 build hours and one upkeep hour a month, break-even is ($2,000 + $960) ÷ $34.60 = 86 sales. Note the fee difference in the same table row: sales that arrive through Gumroad's discovery marketplace are charged 30%, which turns $34.60 into $27.30 and pushes break-even to 109. The distribution channel moves the answer by 23 sales, which is more than most people's marketing moves it.

Templates work better than courses because the build cost is small enough to write off if you are wrong, and because the thing you are selling is usually a by-product of client work you were paid for once already.

The ebook, which is a marketing asset misfiled as a revenue line

766 copies is the number once the half-hour a month of upkeep is counted alongside the build. That is a genuinely hard number for a self-published niche title with no distribution engine behind it. Note also that the 70% band on Amazon.com is $2.99 to $12.99, expanded from $2.99–$9.99 on 7 July 2026; price outside that band and you drop to 35%, at which point the same $9.99 book nets $3.50 and break-even moves past 1,500 copies.

Write the book because it wins you consulting work, or because it makes you the person people quote. Do not write it as a revenue plan.

The paid newsletter, where the flat fee bites

Substack takes 10%, and Stripe adds 2.9% + $0.30 per transaction plus a 0.7% recurring billing fee. On an $8 monthly subscription, that $0.30 flat fee is 3.75% on its own, and billed monthly it recurs twelve times: 12 × ($0.232 + $0.30) = $6.38 a year, against $9.60 for Substack's cut and $0.67 for recurring billing. Gross $96, net $79.35. Bill the same subscriber annually at $80 and you pay the flat fee once instead of twelve times, which is worth roughly $3.30 a subscriber a year — small, until you have 400 of them.

Six hours a month of writing is $5,760 of forgone billing a year, so you need 73 paying subscribers just to break even against the hours, held for the whole year, replacing churn as it happens. Unlike the other four, this category has no finish line. Stop publishing and the revenue stops.

The stock library, where the honest answer is that you can't model it

Adobe Stock publishes the royalty share, 33% on photos, vectors and illustrations and 35% on video, but the price paid per download depends on the buyer's subscription plan and applicable discounts. That means the royalty percentage tells you the split without telling you the amount, and you cannot compute a break-even in advance from published figures. Nobody should pretend otherwise.

There is a directional signal, though, and it is not encouraging. Freelance image-creation job posts fell 17% after image generators launched (Demirci, Hannane and Zhu), which measures commissioned work rather than stock licensing, but the buyers overlap. Building a library on a percentage you can't multiply by a price you can't predict, in a category with that headwind, is speculation. It can be a fine hobby. It is not a plan.

The alternative nobody prices: 10% on the rate you already charge

Every product above is competing for hours against your existing business, so compare it against the cheapest available intervention.

Take 20 billable hours a week over 46 working weeks: 920 hours at $80 is $73,600. Raise the rate 10%, to $88, and the same 920 hours produce $80,960. That is $7,360 a year for approximately zero build hours. The course, at 101 sales, has not yet reached zero at that point. The template pack, at 86 sales, has cleared its build and produced nothing beyond it.

The rate rise is not free, and it is dishonest to pretend it is. Some clients say no. If the client representing 15% of your hours leaves, you are at 782 hours × $88 = $68,816, which is $4,784 less than before — until you refill those 138 hours, at which point you are back at $80,960. So a rate rise is really two moves: reprice, then refill. The refill is the risky half, and it is the half worth planning. Which pricing model you are raising matters too, since hourly, day-rate and fixed-price each fail differently under a repricing; hourly, day rate, fixed price, or value: when each one is wrong covers that choice.

Note also that a product adds fixed monthly cost to a business that already carries some. A course platform at $39 to $89 a month sits on top of whatever you pay for operations, and that stack is where the money quietly goes: Bonsai runs $15 to $59 per user per month and HoneyBook $29 to $129 a month at list price. If a new product's platform fee is going on top of a stack you have not audited in a year, audit the stack first — a Bonsai alternative that covers the money side in one workspace may free up the product's entire platform budget before you build anything.

Where the answer flips

The rate rise wins by default, so a product has to beat it explicitly. Using the course figures, the product must clear the $7,360 the rate rise would have produced, plus the build amortized over two years ($9,600 ÷ 2 = $4,800), plus $3,840 of upkeep and $468 of platform: $16,468 a year, or $16,468 ÷ $137.83 = 120 sales a year, every year. Ten a month, sustained.

Three conditions flip the decision, and they are conditions rather than preferences.

You already have distribution. Not "I post sometimes," but an audience or traffic source that reliably puts an offer in front of roughly a thousand qualified people a month. If you have it, the conversion problem is solved and the arithmetic above is achievable. If you do not, building the product means also building the audience, and that build is larger by an order of magnitude.

Your hours are sold out and already repriced. At capacity, with a rise already tested, the rate lever is exhausted and a product is the only remaining path to more income. That situation is real, and much rarer than the number of people building courses would suggest.

The product's job is lead generation rather than revenue. Then judge it on booked project revenue and the break-even changes completely: one $12,000 engagement traced to the ebook repays 60 build hours twice over. Decide that at the start, or you will file a successful marketing asset as a failed product.

If none of those three describe you, raise the rate. It takes an afternoon and one uncomfortable email, and the payoff arrives in the next invoice rather than in year two.

The number to check before you build anything

Before committing the first build hour, price them. Multiply your realistic build estimate by 1.5, multiply that by your effective hourly rate rather than your quoted one, and compare the result against 10% of last year's billings. If the product cannot beat that inside twenty-four months on a conversion rate you can defend, the hours belong in client work and the rate belongs 10% higher.

Worklyn's Rate Check in CFO Mode reports what each finished project actually returned per hour, which is the number any product has to beat before the build hours are worth spending.