Choosing a revenue shape
Subscription, one-time, advertising, marketplace — each shape assumes a different kind of user behavior, and the wrong one guarantees churn no product fix will cure.
You now know, for at least one project, what it costs to keep alive and whether somebody is already charging for something like it. The next question is not how much to charge. It is what shape the money arrives in — because the shape decides how many customers you need, how often you have to find them again, and how much of your life the thing consumes after it launches.
Get the shape wrong and no amount of pricing skill rescues it. A shape that needs two million monthly pageviews is not a pricing problem for a finished sports-guides site with no traffic. It is a category error, and it is much cheaper to catch here than after six more months of building.
The seven shapes
There are more than seven ways to be paid, but these seven cover almost everything a solo builder can actually reach. Read the table for the two columns that matter most at your scale: what it costs to hold steady, and how big it has to get.
| Shape | Who pays, and how often | Maintenance load | What reaching the target takes |
|---|---|---|---|
| Subscription | The user, every month, until they stop | High and permanent — every month you have to be worth it again | Hundreds of customers who don’t leave |
| One-time purchase | The user, once | Moderate — no churn to fight, but nothing compounds either | A fresh crop of new buyers every month, forever |
| Advertising | Not the user — an advertiser, per thousand views | High — the content is the product, and it goes stale | Millions of pageviews a month, at the median rate below |
| Marketplace | Both sides, per transaction | Very high — trust, disputes, fraud, keeping supply alive | Liquidity on both sides before either side is worth anything |
| Services | A client, per engagement or per hour | Nearly none — there is no product to keep alive | A handful of clients — and it stops the day you do |
| Physical goods | A buyer, per object | High and physical — inventory, shipping, breakage, returns | Enough units that a thin per-unit margin adds up |
| Licensing | Another company, per contract | Low — you maintain one thing, somebody else distributes it | Very few payers, each one found and closed by hand over months |
Three of these deserve a sentence of warning beyond what fits in a cell.
Marketplaces are the trap for engineers, because the software is the easy half. The hard half is a cold start: neither side shows up for an empty room, so you hand-build one side before the other will pay attention, and that work is sales, not code.
Physical goods carry a cost that never goes to zero. A small-batch 3D-printing idea has material, machine time, packaging and postage in every single unit, plus the returns you did not budget for. Software’s marginal cost falls towards nothing as volume rises; this does not, which changes the entire shape of the arithmetic.
Licensing has the best maintenance profile on the table and the worst discovery problem. You are not selling to users, you are selling to one buyer at a company, on their procurement timeline, and there are perhaps a dozen plausible buyers in the world for any given thing.
The ad-revenue reality check
Advertising gets its own section because it is the shape people reach for when a project already exists and has no obvious buyer. It looks free: you built the thing, you switch ads on, money appears. The arithmetic below is the reason that almost never works, and it is worth doing before you spend an evening integrating an ad script.
Who is allowed in
Three networks matter for a small site, and each publishes its own entry bar:
- Google AdSense publishes no traffic minimum at all. Its stated eligibility is original high-quality content that attracts an audience, policy compliance, being at least 18, and having access to your site’s HTML — per Google’s own AdSense eligibility page. No pageview or session threshold appears anywhere on it.
- Journey by Mediavine, the on-ramp tier, requires at least 1,000 premium sessions over a 30-day period, per Mediavine’s own Journey help centre. “Premium sessions” is Mediavine’s own unit, not a plain analytics session — do not assume they are the same number.
- Raptive requires 25,000 pageviews a month, plus a geography gate (half your traffic from the US, UK, Canada, New Zealand or Australia at that level) and a domain at least six months old, per Raptive’s own eligibility page.
The one that breaks most people’s mental model: Mediavine’s main network is no longer gated on traffic at all. It runs seven programs, and only the Journey on-ramp has a traffic threshold. Every other tier is gated on the previous calendar year’s ad revenue, starting at $5,000 a year to reach the Official tier and rising through tiers at $100,000, $250,000, $500,000 and $1,000,000, per Mediavine’s own programs page (effective January 2026). That is a harder gate than a traffic number for a hobby project, because it presumes you are already earning ad revenue somewhere else.
Every number here is published by a vendor
There is no neutral registry of ad-network minimums. Each figure above comes from the network’s own website — these are companies publishing the rules that gate their own business, and they move those rules when it suits them. Raptive cut its own entry requirement to a quarter of what it had been in October 2025, and Mediavine replaced a traffic threshold with a revenue threshold in January 2026. Both changes point the same way: networks competing for supply. Treat a published minimum as a current sales position, not a law, and check the page before you plan around it.
Mediavine’s move is worth one extra beat, because the new gate is a number only the network can see. You cannot self-assess against it the way you could against a pageview count.
What a thousand views is actually worth
RPM — revenue per thousand — is the number the whole shape rests on, and there is no independent, audited benchmark for it anywhere. The best-sourced figures available come from Google’s own published AdSense earnings estimator, which gives an implied page RPM per content category and region:
- Full range: $0.67 to $13.33 page RPM across every category and region it publishes.
- Median: $2.33.
- Typical working band: roughly $1.00 to $5.17, once the single best outlier cell is set aside.
The spread is the finding, not the average. End to end it is a factor of twenty — the same traffic, in the worst category-and-region pairing versus the best, is the difference between a real income and a rounding error. Google attaches its own disclaimer: the estimates are not a guarantee or a commitment, and actual revenue depends on advertiser demand, user location, user device, content vertical, seasonality, ad format and size, the number of ads, and currency exchange rates.
The RPM numbers you have probably seen are anecdote
Mediavine and Raptive publish no average RPM at all. The “$15 to $40 RPM” ranges that circulate for managed networks appear only on affiliate-motivated blogs restating publisher-reported anecdotes — pages that earn a commission when you sign up through them. No primary source for them exists. The one managed-network figure with a real source behind it has to be derived: Raptive’s RPM-guarantee eligibility page requires a site to average 100,000 monthly pageviews and have earned $20,000+ in net ad revenue over twelve months, which implies a floor of $16.67 page RPM for a guarantee-eligible site. That is arithmetic from two of Raptive’s own thresholds, not a published average. Label it that way whenever you use it.
The arithmetic
The formula is trivial: monthly pageviews = ($5,000 ÷ RPM) × 1,000. Run the target through the three defensible rates:
- At the median $2.33 page RPM: about 2.1 million pageviews a month.
- At $5.17 — the best realistic niche-and-geography pairing Google publishes: about 1 million pageviews a month.
- At the derived $16.67 premium managed-network floor: about 300,000 pageviews a month.
That last figure cross-checks against Raptive’s own numbers from a second direction. A site at exactly 100,000 pageviews a month earning exactly $20,000 a year is earning $1,667 a month; reaching $5,000 at the same rate needs three times the traffic, which is the same 300,000. Two routes through the vendor’s own published thresholds land on the same answer.
Two honesty caveats to carry with those figures. They are pageviews, not sessions: session-based rates run higher than pageview rates by an amount nobody publishes, and vendors mix the two units freely — Raptive defines page RPM on pageviews but computes its own guarantee on sessions. And every input is vendor-published, including Google’s, which is a category of source you should size an order of magnitude with and never build a forecast on.
When a platform takes a cut
Subscription and one-time shapes shipped through an app store arrive with a third party already holding a hand out, and the durable structure is worth knowing even though the details are in motion.
Apple. The standard split gives you 70% of a subscription in the first year, rising to 85% after a subscriber accumulates a year of paid service, per Apple’s own subscriptions page. The App Store Small Business Program drops the commission to 15% on paid apps and in-app purchases for developers who made up to $1 million in proceeds in the prior calendar year, and for developers new to the store, per Apple’s own program page. Note the trap in that threshold: Apple defines proceeds as sales “net of Apple’s commission and certain taxes and adjustments” — so it is measured after the cut, and gross customer spend is meaningfully higher than $1 million by the time you cross it.
Google Play. The long-standing structure is 15% on the first $1 million of annual developer revenue and 30% above it, with auto-renewing subscriptions at 15% regardless of revenue, per Google’s own service-fee page. That same page now carries a different rate card for the EEA, the UK and the US, effective June 30, 2026: 10% plus a 5% billing fee on the first $1 million for new installs and for auto-renewing subscriptions, with higher rates for transactions from existing installs.
This area is actively moving — re-check it
Both stores are mid-restructure under court order as of August 2026, so neither “30%” nor “15%” is a settled fact right now. Apple currently charges 0% on purchases reached through external links in US apps, following litigation; on August 14, 2026 it filed a proposal with the district court for a commission on those link-out sales, which is pending and not in effect. That litigation status comes from press reporting, not from an Apple page — Apple has published no rate. Google’s new EEA/UK/US card is on Google’s own page, but the rollout beyond those markets is reported rather than published.
Plan on the durable structure — roughly 15% from small developers, roughly 30% from large ones, with the small-developer rate gated on a $1 million annual threshold — and re-read both vendors’ pages before you act on this. If you are reading this six months from now, at least one number above has changed.
Everyone in this section is describing their own toll booth. Apple and Google publish the commission rules for the stores they own; Google also publishes the AdSense estimates above and, on its own revenue-share page, states that publishers receive 80% of revenue after the advertiser platform’s fee, or about 68% when advertisers buy through Google Ads. None of these are neutral disclosures. They are all accurate as far as they go, and all of them stop exactly where it suits the publisher to stop.
Why services win on economics and lose on leverage
Look back at the services row. It has the best numbers on the table by some distance: a client pays real money in week one, you need a handful of them rather than hundreds, there is no acquisition funnel, and the maintenance load when you stop is zero because there is no product to keep running. If the only question were “what reaches $5,000 a month fastest”, services would win and the lesson would end here.
The reason it does not is the thing Nathan Barry’s essay is about. Barry models income not as one ladder but as several separate ladders — trading time for money as an employee, running your own services business, productizing those services into fixed-scope and fixed-price offerings, and selling products — where you climb rungs within a ladder and also step across to a more advanced one. Earnings rise in both directions. So does the difficulty: “the difficulty increases with each move as well” (Nathan Barry, The ladders of wealth creation, 2019).
That is the actual axis of his essay — earning potential against difficulty and skill required. It is not a leverage-versus- reliability trade-off, and “audience” is not one of his ladders; audience is the asset he describes building underneath the product ladder, not a rung on it. Worth stating plainly because the essay gets misquoted in both of those directions constantly.
The reliability half of the argument is this course’s own extension, not Barry’s wording, and it is the one that decides the shape. Services income is the most reliable money on the table and it stops the day you stop. It does not compound, it cannot be sold, and it consumes exactly the evenings a product would need. A services partnership can be the correct answer for a year — funding the time to build something with leverage is a legitimate use of the lowest-leverage shape. It is only a mistake when it is chosen by default and never revisited.
Retrieval check
Without scrolling up: Barry’s essay is often summarised as a trade-off between leverage and reliability across services, products and audience. Two things in that summary are wrong. What are they, and what does he actually say?
Check your answer
First, the axis. Barry’s stated trade-off is earning potential against difficulty — earnings rise as you climb a ladder or move to a more advanced one, and so does the difficulty of each move. Reliability of income is a reasonable implication to draw, but it is not his framing, so it should not be attributed to him.
Second, the taxonomy. Audience is not one of his ladders. The four are trading time for money, your own services business, productized services, and selling products. Audience is the asset he says you build underneath the product ladder.
Now apply it: pick a shape, and reject one out loud
Hands on
Choose a revenue shape for one project
Done when: PORTFOLIO.md’s Candidate revenue shape field holds one shape for one project, plus at least one shape you rejected and the reason.
- Take the project you already measured a maintenance number for. Not a new one — the shape decision is only useful against costs you have actually counted.
- Write down who the payer is under each of the seven shapes, in one line each. Several will be obviously impossible for this project. That is the point: eliminating is faster than choosing.
- For every shape still standing, write the scale it demands — customers, buyers, pageviews, units, contracts. If advertising is still on your list, do the arithmetic explicitly:
($5,000 ÷ RPM) × 1,000, using a rate from the band above that matches your actual subject matter and audience geography, not the best cell in the table. - Compare each surviving shape’s demanded scale against the traffic or customer count the project has today. Most shapes die here, and the ones that die loudest are the ones worth writing down.
- Write the winner into
PORTFOLIO.mdunder Candidate revenue shape for this project. Add one rejected shape and one sentence on why it lost — the rejection is what stops you quietly re-adopting it in three months. - Bring both into the chat. I’ll push hardest on an advertising answer, since it is the shape whose arithmetic people skip, and on a services answer that has no stated end date attached to it.
Check your recall
Answer from memory — no scrolling back.
What this does not cover
You have a shape and a rejected alternative, which is a decision about the kind of money. It is not yet a number. Next comes the arithmetic that does the actual killing: the target divided by a price you could realistically charge, giving the customer count you need — and then the churn multiplier that turns that count into a treadmill, because a shape where customers leave every month means the number is not how many you need, but how many you need to find again, every month, forever.
Read this next — primary source
The ladders of wealth creationNathan Barry, nathanbarry.com, December 3, 2019 — free, about 15 minutes
Barry models income as several separate ladders rather than one — trading time for money, running your own services business, productizing those services, then selling products — and the trade-off he actually states is earnings against difficulty: “the difficulty increases with each move as well.” Read it for that axis, and read it knowing whose essay it is: Barry founded the email-marketing company Kit (formerly ConvertKit), the essay uses that company’s growth as its worked example, and the ladders he ranks highest are the ones his company sells tooling for.
Stuck, curious, or think this lesson is wrong? Ask your teaching agent. The lessons are the scaffold; the conversation is where the learning gets unstuck.