Is Subscription Pricing Still Right for Your SaaS, or Should You Switch to Usage-Based?
Neither wins by default: the right model depends on whether your costs scale with usage, and for AI features, they almost always do.
Key takeaways
- Subscriptions win on predictable revenue and simple billing; usage-based wins on aligning price with cost, especially when a feature has variable marginal cost like AI inference.
- Hybrid models (base subscription plus usage overages) are increasingly common precisely because AI features break the flat-fee assumption.
- The core question isn't "which model is better" but "does my marginal cost per customer vary a lot, and can my billing system track it."
- Usage-based pricing without cost visibility is dangerous: you can undercharge for heavy users and not notice until margin erodes.
- Model your worst-case heavy user cost before choosing a billing model, not after.
Why is subscription pricing being questioned now?
An Indie Hackers discussion asked whether flat subscription pricing still makes sense given that many products now meter usage by API calls, storage, or AI usage under the hood. The argument: subscriptions assume the cost of serving each customer is roughly flat, but AI features (LLM calls, image generation, embeddings) have real, variable marginal costs that scale with how much a customer actually uses the product. A flat subscription hides that variance, meaning some customers are wildly profitable and others lose money on every login.
What does usage-based pricing actually solve?
Usage-based billing ties price to consumption, so cost and revenue move together. If a customer runs 10x the AI workload of another, they pay roughly 10x more (or some scaled amount), instead of both paying the same subscription fee while one silently drains your margin. This matters most for founders whose core feature has a real per-unit cost, tokens processed, images generated, minutes transcribed, rather than a mostly fixed cost like storage or seats.
The tradeoff: usage-based pricing is harder to forecast for both you and your customer, adds billing complexity, and can create sticker shock if a customer's usage spikes unexpectedly.
When does a hybrid model make more sense?
Most SaaS products with an AI layer are converging on a hybrid: a base subscription covering the predictable, low-variable-cost parts of the product, plus metered usage or overage charges for the parts with real marginal cost. Say a base plan includes 1,000 AI-assisted actions per month, and usage beyond that is billed per action at a rate that covers your token cost plus margin. This keeps billing simple for typical customers while protecting margin against outlier usage.
How do you decide without guessing?
Model your cost per unit of usage first: what does one AI-assisted action actually cost you in tokens, and at what usage level does a flat-fee customer become unprofitable? If that breakeven point is comfortably above your typical customer's usage, a subscription is fine. If heavy users regularly cross it, you need metering, a hybrid tier, or usage caps with clear overage pricing.
The takeaway
There's no universally "best" model, only the model that matches your actual cost structure. For products with AI features carrying real marginal cost, a hybrid of subscription plus metered usage is usually the safer default. Calcaas helps you model tiers, packaging, and break-even usage levels before you commit to a billing structure.
Frequently asked questions
Is usage-based pricing better than subscriptions for AI products?
Not universally better, but it's usually a better fit when the AI feature has real per-unit cost (tokens, generations, minutes) that varies significantly across customers. Products with mostly fixed costs can stick with flat subscriptions.
What's the risk of getting this wrong?
Charging a flat fee for a variable-cost feature means heavy users can be unprofitable without you noticing until margin erodes across the customer base, often for months before it shows up in reporting.
How do I set a fair overage rate?
Start from your actual cost per unit (token cost per action, for example), add your target margin, then round to a customer-friendly price point. Recalculate whenever provider pricing changes.
Can I switch from subscription to hybrid without upsetting existing customers?
Yes, most successful transitions grandfather existing customers on their current plan for a period while offering the new hybrid structure to new signups, then migrate gradually with advance notice. (Note: place this JSON-LD inside a <script type="application/ld+json"> tag in the page head.)
More from the blog
Pricing StrategyWhy Hasn't SaaS Pricing Raced to the Bottom? The Uncomfortable Answer for Founders
It's not because competition is weak, it's because price is rarely the thing customers are actually optimizing for, and founders who compete on price alone usually lose anyway.
Pricing StrategyToken Volume Is Up 25x and Mid-Tier Models Do 90% of the Job. Your Pricing Has to Move.
Tom's Hardware reports token volume exploding 25-fold while mid-tier models deliver roughly 90% of flagship capability at one-sixth the cost, which means the default choice of flagship-everything is now a margin decision rather than a quality one.
Pricing StrategyA Frontier Model Launch Is a Repricing Event: A Six-Step Checklist for Founders
When a provider ships a new flagship at a new rate, your cost of goods sold changes before your pricing page does, so treat launch day as a margin review rather than a feature announcement.
Pricing math, in your inbox.
One short note a week on AI pricing, token economics, and margin. No spam, unsubscribe anytime.