Usage-based pricing: when it beats a retainer

Usage-based pricing beats a retainer when the amount of work or cost varies a lot from month to month and neither side can predict it: AI API costs, content volume, support tickets or ad spend. It loses when the value of the work does not depend on how much of it you deliver. The safe form is a floor, a unit price above the floor and a cap.

Software made usage-based pricing normal

Clients have been trained to expect it. A January 2025 survey by Metronome and Greyhound Capital of 100 SaaS companies found that 85% had adopted usage-based pricing in some form. That is a small, vendor-run sample, so treat it as a sign of the direction, not a market census.

The more useful finding for service businesses is the mix. Maxio's 2025 Pricing Trends Report, based on a Benchmarkit survey, says companies using hybrid models (a subscription plus usage) reported the highest median growth rate at 21%. It also says 73% of SaaS companies with usage-based models actively forecast variable revenue. In plain terms: the businesses that did well kept a base fee and added a usage part, and the ones that went fully variable had to plan for the swings.

That is the pattern to copy. Pure usage means a zero month when usage drops to zero. Pure retainer means you absorb the months when usage is high.

The three shapes it takes

Stripe's pricing model reference and its usage-based billing docs list the common usage models as fixed fee and overage, pay as you go, and credit burndown. For service work they look like this:

1. AI cost pass-through. You charge a base fee for your work and pass on the real AI usage cost with a handling margin.
2. Per-deliverable pricing. A set price per article, per asset or per merged feature, billed monthly.
3. Volume tiers on a retainer. A fixed monthly fee covers up to N units (tickets, posts, requests), and each unit above N costs a set amount.

The third is the closest to metered billing in software, and the one most solo owners can run without special tooling.

Where usage-based pricing wins

It wins when the quantity is the real variable. Five common cases:

• AI-heavy build or automation work. Model costs are billed by the amount of text processed, and a model provider's price list such as Claude's pricing page quotes a separate price per million tokens for input and for output. A project that runs many agent loops can cost several times what a simple one does. Pass-through gives the cost to the side that controls the usage.
• Support and moderation retainers, where ticket volume follows the client's sales, not your effort.
• Content production at scale, where the client's real question is "what did I get for what I paid?" and a per-piece price answers it.
• Ad spend management, where a percentage of managed spend rises and falls with the budget.
• Automation and agent work billed per run or per task, because running costs are uneven.

Where it is the wrong call

Skip usage pricing when the quantity is not what the client values.

• Brand identity, a logo or a campaign concept. Price per deliverable misreads the value, which is strategic and lasts for years. Quote a project price based on the outcome.
• Advisory work. A monthly strategy call is paid for judgment, not throughput. A flat retainer is better.
• A fixed deliverable with a hard date, like a website launch. Fixed price aligns both sides on scope and deadline.

If you charge one of these by the unit, the client starts counting units and you start doing less, and neither of you is thinking about the result.

How to set it up without a billing fight

Most disputes come from vague units and surprise invoices. Six rules prevent them:

1. Define the unit in writing. "One published article of 800 to 1,200 words with one round of edits" is a unit. "An article" is an argument.
2. Set a floor. A minimum monthly fee covers your availability even in a quiet month.
3. Set a cap. A client who cannot see the maximum bill will fear the worst. A cap of, say, 2.5 times the floor is a starting point to negotiate, not a rule.
4. Bill monthly, not per event. Dozens of small invoices create reconciliation work for both of you.
5. Show usage as it happens. A simple monthly report or a message each time usage passes a threshold prevents invoice shock.
6. Review the unit price every quarter. Usage patterns change, and a price set in January may be wrong by July.

Our guide to retainer pricing covers how to set the floor, and the retainer vs project vs hybrid comparison shows how agencies mix them.

The AI cost pass-through pattern in detail

When AI is a meaningful share of the work, the structure that works is three lines on the agreement:

• A base fee for your time, judgment and review.
• The actual AI bill plus a handling margin (many people use 15% to 25%), invoiced monthly with the receipts attached.
• A monthly ceiling on AI spend, so a runaway loop cannot surprise the client.

Here is the arithmetic. A $25,000 build uses $1,500 of model costs. At a 20% handling margin, the client pays $1,800 on top of the base fee and sees the receipts that support it. You are not hiding the cost in your rate and not inflating usage to raise the bill. If you would rather not itemize, fold a range into the base fee and say so.

Do not pass through anything you cannot document. If the client asks to see a receipt and you cannot show one, the margin becomes a trust problem.

A matrix you can use

Pick by what is variable on this job:

• Hourly: discovery and unclear scope, where the work is the unknown.
• Fixed price: a defined deliverable with a deadline.
• Flat retainer: ongoing advice or steady work at predictable volume.
• Usage-based: AI costs, per-piece content, volume-driven support.
• Hybrid: a base fee plus usage above a threshold, which is the most common form in practice.

Use more than one model across your clients. Forcing every client into the same one is how you underprice the unpredictable jobs.

How Delivvo bills a floor, an overage and a cap

Delivvo does not meter usage for you. You read the numbers each month and enter them yourself. What it does is keep the agreement, the floor and the monthly bill in one thread, so the structure is clear on both sides.

1. Build the offer in proposals: the floor as a package, extra units as add-ons the client can see and accept.
2. Put the unit definition, the cap and the quarterly review in the contract. Signing is typed or drawn, and the signing record is archived with the PDF.
3. Bill the floor with a recurring invoice in your own number format.
4. At month end, send a second invoice for the overage or pass-through, with the receipts attached as files in the client portal. The client opens the portal with a one-time code, no password.
5. If the engagement needs staged payments, a payment schedule can split a fee into up to 12 payments.

A detail that fits this model: invoices carry your tax details, support reverse charge, and every invoice has a verification code. Clients pay through your own gateway, and Delivvo takes 0%, so your overage is not shared with a platform. Check tax wording with your accountant.

The cap works the same way: write the ceiling into the contract, and when usage nears it, message the client in the portal before sending the invoice, so the conversation about raising it happens before the bill, not after.

See the recurring invoices page for how the monthly cycle works.

For the wider picture, see quotes and proposals and the CRM for service businesses.

FAQ