Founders ask Collective 54 this 8 times in our records, 2 of them in 2026. Three different models are bundled into one question, and they fail for different reasons.
A productized service is a defined scope sold at a fixed price, repeatedly, with the same inputs and the same deliverable. The client buys a known thing.
A subscription is paid to gain access to an asset or a continuing output. Clients want access to proprietary data, to a methodology, to a stream of work, to you. The client buys ongoing availability, and the proposal cycle shrinks because the value sits in the availability itself.
Usage-based pricing charges according to consumption. The client buys a rate.
These are adjacent on a pricing page and completely different underneath. A productized service lives or dies on scope discipline. A subscription lives or dies on retention. Usage-based pricing lives or dies on the shape of your marginal cost curve.
This is the decision everything else depends on: what does the price scale with.
The rule that makes the choice for you is that packaging must decouple from labor, because clients do not value effort, they value progress. Across the three eras the pattern is consistent. Era 1 moves from hours to access. Era 2 moves from deliverables to results. Era 3 moves from subscriptions to outcomes. The less your pricing is tethered to time and headcount, the more scalable the model becomes, and time is the least defensible value metric in a world of automation.
So test any candidate metric with one question. If you get twice as fast at delivering this, does the price fall? If yes, you have picked a metric that measures your effort, and you have built a model that punishes you for improving.
The related test is whether you can quantify what the client receives in hard dollars, and explain the logic of the price in a way that makes sense to them. Most boutiques price badly because they do not know what their services are worth to their clients, cannot quantify the value received, and price inward out from internal cost. That failure hurts twice as much in a recurring model, because the mistake renews every month.
Before you price one, confirm you have something to sell continuing access to.
Subscriptions and licensing both rest on an asset the client wants access to: proprietary data, a methodology, a tool, a body of work that exists whether or not a project is running. If nothing exists between engagements, what you are calling a subscription is a retainer, and it should be priced and structured as one.
This matters beyond pricing. An asset the firm owns and clients pay for access to is intellectual property in the sense a buyer cares about, and it is the difference between a firm that sells the same thing many times and one that sells its people repeatedly.
Recurring pricing fixes your revenue per client and leaves your cost floating, which is exactly the wrong way round if you have not modeled the cost.
Forecast cost to serve across human, AI and tooling components, and model expected and acceptable margins before signing rather than discovering them afterward. Then sensitivity test the model against discounting and scope creep, because those are the two forces that will move it.
Era 3 delivery mixes human judgment with AI execution, which changes the shape of the problem. You need continuous clarity on marginal cost behavior, on how human and machine effort are allocated, on how variable delivery is by engagement type, and on how AI usage changes your cost curves over time. In a project business a bad cost assumption costs you one engagement. In a subscription it compounds every renewal.
Usage-based pricing is the model most likely to be adopted for the wrong reason, which is that it feels fair.
It works when your marginal cost is genuinely linear in the unit you are charging for. It fails when it is not, and in AI-assisted delivery it frequently is not. Cost does not scale smoothly with volume, and it can move materially between quarters as models, pricing and usage change. If your unit price was set against the cost curve of an earlier quarter, a heavy user can be unprofitable at a price that looks healthy on average.
The second problem is behavioral. Usage-based pricing gives clients an incentive to consume less of the thing you want them using more of, which is a strange position for a firm whose value depends on being embedded.
Use it where consumption genuinely tracks cost and where more usage is a good sign. Otherwise put a usage component on top of a subscription floor rather than making usage the whole model.
Versioning lets clients choose their own price, which makes them decide faster and links price to value in their own thinking. Tier on responsiveness, personalization or complexity rather than on volume of work, and charge the most for the features clients want most while charging least for the ones they do not care about.
Then enforce the boundaries, because this is where recurring models actually break. Most pricing breakdowns happen through packaging drift rather than headline price: exceptions become norms, customization creeps in, and consistency in what is included quietly disappears. Watch the frequency of exceptions, not the merits of each one, and keep a single source of pricing truth. A price that is not enforced is not a price.
Watch for the alignment break as well. When delivery speed increases materially, when AI substitutes for human labor, or when scope expands without a price adjustment, the pricing logic no longer reflects how the work is done. That is how firms scale inefficiency invisibly.
If your work is genuinely bespoke each time, do not productize it to get a cleaner pricing page. Fixed price on undefined scope is how firms lose money quietly, and the discipline you need first is scoping, not packaging.
If you have no asset that exists between engagements, you do not have a subscription and should stop trying to price one.
And if your delivery is not yet consistent, skip outcome-based pricing for now. Outcome pricing becomes feasible when delivery is consistent, impact is trackable and the offer is tightly scoped. Reaching for it early transfers a risk you cannot yet control onto your own margin.
Separate the three. Productized services are fixed scope at a fixed price and depend on scope discipline. Subscriptions sell access to an asset or a continuing output and depend on retention, and if nothing exists between engagements you have a retainer rather than a subscription. Usage pricing sells a rate and depends on marginal cost being linear in the unit you charge for, which in AI-assisted delivery it often is not. For all three, choose a value metric that scales with client progress rather than your effort, testing it by asking whether the price falls if you get twice as fast, since packaging has to decouple from labor and time is the least defensible metric in a world of automation. Model cost to serve across human, AI and tooling before you commit, sensitivity test it against discounting and scope creep, and keep watching marginal cost behavior because a bad assumption compounds at every renewal. Tier on responsiveness, personalization or complexity so the client picks their own price, then hold the boundaries, since most recurring models break through packaging drift rather than through the headline number.
Yes, and treating them as one is the first mistake. A productized service is a defined scope sold repeatedly at a fixed price, and it lives or dies on scope discipline. A subscription is paid to gain access to an asset or a continuing output stream, and it lives or dies on retention. Usage-based pricing charges by consumption, and it lives or dies on whether your marginal cost is actually linear in the unit you are charging for. They look adjacent on a pricing page and behave completely differently underneath, so the failure modes have nothing in common.
Pick something that scales with client progress rather than with your effort, because packaging has to decouple from labor and clients do not value effort. The evolution runs from hours to access in Era 1, deliverables to results in Era 2, and subscriptions to outcomes in Era 3, and time is the least defensible value metric in a world of automation. Test a candidate by asking whether the price falls if you get twice as fast at delivering. If it does, the metric measures your effort and the model punishes you for improving.
That you fixed revenue per client and left cost floating. Forecast cost to serve across human, AI and tooling components and model expected and acceptable margins before commitment rather than discovering them afterward, then sensitivity test against discounting and scope creep. Era 3 delivery mixes human judgment with AI execution, so you need continuous clarity on marginal cost behavior, on human versus machine effort allocation, on delivery variability by engagement type, and on how AI usage moves the cost curve. In a project business a bad assumption costs one engagement. In a subscription it compounds at every renewal.
Packaging drift. Exceptions become norms, customization creeps in, and consistency in what is included disappears, which is how most pricing breakdowns happen rather than through the headline price. The defenses are watching how often exceptions occur instead of judging each on its merits, and keeping a single source of pricing truth, because a price that is not enforced is not a price. Watch the alignment break too: when delivery speed increases materially, AI substitutes for human labor, or scope expands without a price change, the pricing logic no longer reflects how the work is done.
Sources: Greg Alexander, AI Pricing Strategy, Collective 54, for the meta-principle that packaging must decouple from labor because clients value progress rather than effort, with the packaging evolution running hours to access in Era 1, deliverables to results in Era 2 and subscriptions to outcomes in Era 3; for the finding that time is the least defensible value metric in a world of automation and that labor-based packaging means growth requires more people, margins shrink and delivery becomes unpredictable; for competitively priced subscriptions in which clients prefer continuous access to project-by-project pricing, proposal cycles shrink because value is baked into availability, and cash flow stabilizes; for premium subscriptions tiered on responsiveness, personalization or complexity; and for outcome-based pricing becoming feasible only when delivery is consistent, impact is trackable and the offer is tightly scoped. Greg Alexander, The AI-Native Boutique Firm (Advantage Books, January 2027), specifically The AI Pricing Manager for unit economics clarity covering marginal cost behavior, human versus machine effort allocation, delivery variability by engagement type and how AI usage changes cost curves over time; for packaging discipline, the finding that most pricing breakdowns occur through packaging drift rather than headline prices, and the need for alerts when exceptions become norms; for price integrity enforcement, including the position that a price which is not enforced is not a price and the requirement for a single source of pricing truth; for delivery and pricing alignment checks that detect material increases in delivery speed, AI substitution for human labor and scope expansion without price adjustment, preventing firms from scaling inefficiency invisibly; and for modeling expected and acceptable margins before commitment with sensitivity testing against discounting and scope creep. Greg Alexander, The Boutique: How to Start, Scale, and Sell a Professional Services Firm (Advantage, 2020), chapter 4 for subscriptions as payment for access to an asset such as proprietary data and for licensing fees paid for the right to use a methodology or tool; chapter 15 for the reasons boutiques misprice, including not knowing what services are worth to clients, being unable to quantify value received and pricing inward out from internal costs, and for price versioning that lets clients choose their own price while charging most for the features clients want most.
Collective 54 is the private community for founders and executives of boutique professional services firms between $5M and $50M in revenue. Members work these answers against their own numbers.