Shopify runs two large revenue streams inside one company, and its own annual report tells you they are not equally good. Merchant Solutions, driven largely by Shopify Payments, is the bigger growth engine. It also carries lower gross margins than Subscription Solutions because of third-party costs, and the company states plainly that the continued growth of merchant solutions has caused, and may continue to cause, a decline in its overall gross-margin percentage. The company grew revenue 30% to $11.5 billion in 2025 on $378 billion of GMV, and still had to warn investors that its mix was pulling margin down. That single disclosure is the cleanest public proof of why some revenue streams are more valuable than others, even when they sit under the same roof, serve the same customers, and get added together on the same income statement.
Revenue is the most-watched number in small business and the least informative. It tells you what arrived. It tells you nothing about what it cost to make it arrive, whether it will arrive again, or what someone else would pay for the right to receive it.
The second price tag on every dollar
Every revenue stream has a cost to serve that rarely appears in the way owners talk about it. Onboarding time. Support load. Refunds and chargebacks. Payment processing. The senior person who has to get involved when something goes wrong. The working capital tied up before the money lands.
Public filings make this visible in a way most private books never do. ServiceNow reported a 5% gross loss percentage on professional services and other revenue in 2025, worse than the 2% loss in 2024. Dayforce disclosed professional services gross margin of negative 19.9% in 2024, an improvement on negative 22.7% in 2023. These are large, well-run software companies deliberately booking revenue that loses money on delivery.
They do it because that stream is a gateway. Dayforce also disclosed that the share of customers live for more than two years rose from 85% to 87%. Implementation work is the toll you pay to reach years of high-margin recurring revenue behind it. The loss is an acquisition cost wearing the costume of revenue.
Which gives you the first useful distinction. A low-margin stream is worth keeping when it reliably manufactures a high-margin stream, and worth cutting when it only feels productive. The test is whether you can name the conversion rate. If a service line is supposedly a gateway but nobody has ever measured how many of those clients become retainer clients, it is not a gateway. It is just unprofitable work with a story attached.
Before you compare streams on anything else, strip each one down to contribution after the real cost to serve. This is slower and more annoying than it sounds, because shared overhead resists allocation. Do it roughly rather than not at all. A stream that looks like a third of your revenue and turns out to be a twentieth of your profit changes decisions immediately. If you have never done this exercise, start with finding the most profitable part of your business before touching your pricing or your product line.
Durability: the quiet quarter test
Here is a question worth sitting with. If you stopped all selling for one quarter, no outreach, no ads, no proposals, what revenue would still show up?
That number is your durable revenue. Everything above it is revenue you manufacture monthly with effort. Both spend the same, but they are not worth the same, and the gap between them is the single biggest driver of what a business feels like to own.
Durability comes in three tiers, and conflating them causes a lot of bad planning.
Manufactured revenue
It exists because someone sold it this month. Project work, one-off installs, campaign-driven e-commerce spikes. It can be excellent revenue. High margin, fast cash, no long obligation. But it resets to zero on the first of every month and consumes attention permanently. Stripe makes the structural point: without recurring revenue, a business has to keep acquiring new customers or upselling existing ones simply to stay profitable.
Habitual revenue
The customer keeps coming back, but nothing binds them. Repeat purchases, informal retainers, the client who has called every January for six years. Owners treat this as recurring revenue. Buyers do not. Habit is real and valuable, but it is a behavioural pattern, not a commitment, and it tends to break precisely when conditions get difficult.
Committed revenue
Contractual, dated, with defined notice periods and renewal terms. This is what the market means by recurring revenue. Adobe uses annualised recurring revenue as the key metric management relies on to assess the health and trajectory of its Digital Media segment, exiting fiscal 2025 with $25.20 billion of total ARR, up 11.5% year over year. When a company of that size tells regulators that ARR is the number it steers by, it is telling you something about what committed revenue does to a business.
The practical move for a smaller business is usually not inventing a new product. It is converting habitual revenue into committed revenue. Same customers, same work, different structure: a twelve-month agreement with a notice period instead of an open-ended arrangement that both sides renew by inertia. That conversion costs almost nothing and changes both your forecasting and your valuation.
One caution, because this is where generic advice goes wrong. Recurring does not automatically mean better. Recurring revenue that is heavily discounted, churns at 4% a month, or carries a support burden that scales one-for-one with the customer count can easily be worth less than project revenue with strong margins and pricing power. Predictability is a multiplier on quality, not a substitute for it.
Concentration, and the diversification that is not
Accounting and securities rules both converged on the same threshold: a single external customer providing 10% or more of revenue has to be disclosed, and under Item 101(c)(1)(vii) of Regulation S-K a company must name any customer at that level if losing them would materially hurt the business. Regulators decided that 10% is the point at which a customer stops being a detail and becomes material information for anyone assessing the company.
Borrow the threshold. It is a better private-company heuristic than anything in the advisory literature, because it was set by people whose job was to decide when a dependency becomes dangerous.
Then extend it past customers. Run the same 10% test on channels, platforms, referral partners, and individual salespeople. A stream where 60% of volume arrives through one marketplace’s algorithm has a concentration problem whether or not any single customer is large.
Which leads to the part almost every article on this topic misses. Adding streams does not reduce risk unless the streams fail for different reasons.
A design studio that sells project work, retainers, and a template shop has three revenue streams and one risk. All three depend on the same marketing channel, the same two senior designers, and the same category of client budget. A downturn in that client segment takes all three down together in the same quarter. The count went up. The exposure did not move.
So the question is not how many streams you have. It is whether you can describe an event that kills one and leaves the others intact. If you cannot, you have one business with three product names. Independence is the thing being bought, not variety, and variety without independence just adds operating complexity, management attention, and capital requirements to a risk profile that has not improved. Before launching anything new, it is worth being deliberate about whether the opportunity actually fits the business or simply looks attractive from a distance.
Can the revenue survive you?
Transferability is the dimension owners underweight most, because it is invisible while they are still there.
Ask who the customer believes they are buying from. If the honest answer is a named individual, the revenue is attached to a person rather than to the company. That revenue is still real, still profitable, and still worth having. It is worth considerably less to anyone else, and it caps how far the business can grow, because the constraint is a calendar.
The same test applies to operational knowledge. If a stream depends on a process that lives in one person’s head, it is a stream with a single point of failure that no org chart will reveal. This is usually the real reason an apparently healthy business cannot scale, and it is worth examining alongside the other structural limits on whether a business model can scale.
Cash timing deserves a brief mention in the same breath. Two streams with identical annual contribution can have opposite effects on the business if one is billed annually upfront and the other is invoiced 45 days after delivery of work you funded with your own payroll. Upfront-billed revenue finances growth. Trailing-paid revenue consumes it. Over a year of expansion, that difference can matter more than several points of gross margin.
Pricing structure is part of revenue quality
How you charge changes what the revenue is worth, not just how much of it there is. McKinsey’s pricing research found that a 1% price increase has more impact on bottom-line margin than a 1% uplift in volume or a 1% reduction in costs. The structure of the price carries even more leverage than the level.
Two cases make the point well. Rolls-Royce’s power-by-the-hour model moved engine economics from one-time sales to payment tied to customer outcomes, and McKinsey reports it produced 9% annual underlying services revenue growth from 2004 to 2011 alongside a more than 30% improvement in average time between engine removals. BASF, under a different pricing model, reduced paint consumption per car by 20% while achieving 20% higher margins and a 40% increase in European market share. In both cases the supplier made more money by helping the customer consume less. That only works when the price is attached to the outcome rather than to the unit.
The current debate about subscriptions versus usage-based pricing has largely been settled in practice, and the answer is hybrid. Stripe’s own guidance frames hybrid pricing as a stable recurring base with bills that scale with usage, and McKinsey has documented software pricing moving away from pure seat-based models toward hybrid and consumption-based monetisation in the AI era. The reason is structural: where the cost of delivery varies with use, a flat subscription either overcharges light users or destroys margin on heavy ones. A floor plus a meter solves both.
Consumption revenue carries its own hazard, though, and it is a new one. OpenAI launched prompt caching in October 2024 with a 50% discount on cached input tokens, applied automatically to prompts longer than 1,024 tokens in 128-token increments, then later reported discounts of up to 90% on cached input with improved caching. If your revenue is metered against someone else’s units, your effective price can fall by most of its value because of an infrastructure decision you had no part in. Usage-based revenue built on a platform you do not control is a stream with a dial that someone else is holding.
What the market actually pays for
The old shorthand said get to ARR and the multiple takes care of itself. The data no longer supports the second half of that sentence. Bessemer reported the average Cloud 100 multiple in 2025 at 20x ARR, down 41% from the 2023 peak, with the BVP Nasdaq Emerging Cloud Index trading around 14x run-rate revenue. Recurring revenue still commands a premium. It simply does not command a fixed one.
Bessemer’s Rule of X work sharpens this further: the weighting the market applies to revenue growth versus free-cash-flow margin has ranged from roughly 2x to 9x over the past five or six years. The identical revenue stream, with identical mechanics, is valued differently depending on which of those two things the market happens to be rewarding that year.
Those multiples belong to public cloud companies and do not transfer to a ten-person consultancy. The logic does. A buyer is not purchasing last year’s revenue. They are purchasing a probability distribution of next year’s, discounted for how much work and risk sits behind it. That is why diligence segments revenue and discounts parts of it: concentrated accounts, uncontracted habit, founder-attached relationships, and streams whose margin depends on a platform policy all get marked down, regardless of what the top line says.
Which is also the clearest statement of what separates revenue growth from growth that is actually worth having, a distinction worth reading about in its own right.
Running the test on your own business
The SBA’s business-plan guidance names revenue streams as a component you are expected to define. Defining them is the easy part. Grading them is the work. Take every stream producing meaningful money and answer six questions for each, separately, with numbers where you have them and honest estimates where you do not.
- Contribution after cost to serve. Not gross revenue. What is left after support, onboarding, refunds, payment costs, and the senior time it consumes.
- Renewal cost. What you must do again next year to earn the same money. A stream that renews itself and a stream you rebuild every January are different assets.
- Commitment. Contractual, habitual, or manufactured. Be strict. Habit is not a contract.
- Concentration. Any customer, channel, platform, or person above 10%.
- Independence. Name the event that kills this stream. Does it kill the others too?
- Transferability. Does it survive you leaving for three months?
You will almost certainly find one stream that is large, visible, exhausting, and contributing far less than its size implies. The instinct is to fix it. Often the better decision is to stop it, which is easier said than done, because cutting revenue feels like failure even when it raises profit. That resistance is why saying no is a growth skill rather than a defensive one.
One last warning against chasing a perfect model. Research by Malone, Weill and colleagues examining whether some business models perform better than others found that no model outperformed on all dimensions of market value, profitability, and operating efficiency. There is no configuration of revenue streams that wins everywhere at once. High margin often comes with low durability. Durable contracts often come with heavy delivery costs. Consumption pricing aligns beautifully with value and forecasts terribly.
The goal is not to find the best revenue stream. It is to know exactly which trade-off each of yours represents, and to stop treating dollars as interchangeable simply because they land in the same account.
