Usage-Based Billing for SaaS: How It Actually Works

Behind every usage-based pricing plan is a usage-based billing system. These systems are the backend of pricing and monetization. Most people don't know how they work and end up with a billing system that makes them do all the work.

That's because usage-based billing isn't just a pricing decision. It's an architecture decision. It requires ingesting raw usage data, applying pricing logic, and generating real-time, accurate invoices. All of them are deeply integrated with your product and infrastructure. This isn't something you bolt on later.

Lago is built for this reality. Open-source, event-driven, API-first. Designed to handle complex billing use cases without creating a dedicated internal billing team.

Of course, usage-based billing isn't new. Companies like Twilio have charged per API call for a long time. But AI has created marginal costs in software, so billing is getting more and more usage-based. That's why choosing the right usage-based billing system is an important decision. It's a data pipeline, a billing engine, and a product decision.

Why usage-based billing really matters

In the early days of a SaaS company, you price based on instinct and precedent. $99/month feels right. Maybe you've got a Pro plan for power users. Or you add a seat-based charge.

But once you have AI features where your cost scales with usage, static pricing starts to break. One power user can literally bankrupt you. Usage-based billing fixes that.

When done right: lower activation friction (no upfront commitment), cost aligned to value (no more subsidizing your highest-usage customers), and revenue expansion that scales without a sales call. When done wrong: audit failures, billing disputes, and an engineering drag that compounds with every pricing change.

If you've been part of any decent-sized company, you know how billing can become a nightmare. The internal team of billing engineers who do nothing but maintain a homegrown system that everyone hates but is too complex to replace? We've been there, done that, hated it and decided to fix it.

Before we dive into the details, let's define usage-based billing so we're on the same page.

What is usage-based billing and how does it differ from tiered pricing?

Usage-based billing charges customers based on what they actually use: $0.01 per API call. Tiered pricing offers fixed plans with more or less unlimited usage. Of course, there are many hybrid models, like plans that include a certain amount of usage before you pay overages or upgrade to a new tier. The key difference: usage-based billing scales continuously with usage, while tiered pricing introduces pricing cliffs and requires upgrades. Many companies use both together.

The real models behind usage-based pricing

There's no single usage-based pricing model. In reality, only infrastructure companies have pure usage-based pricing. Most companies use some type of hybrid model.

You'll often see pay-as-you-go models for elastic workloads ($0.01 per API call), tiered pricing for predictable growth ($0.10 per GB for the first 100GB, then discounted tiers), or volume pricing that makes scaling simple.

AI products tend to go hybrid, combining a flat base fee with variable usage (like $99/month for 500 AI credits). Enterprise deals are usually custom but often include prepaid credits that offer budget control while maintaining usage flexibility.

When you have a billing system that supports all of this, pricing becomes code quickly. When your billing system is a hot mess, it might take months to ship a minor pricing change.

Simulating invoices before going live

With Lago, you can run simulations in staging using real usage data. Create pricing plans, load in historical usage, and generate preview invoices. This lets you answer: "What would this customer have paid last month under the new pricing?" It's the safest way to test new pricing before going live.

How usage-based billing systems actually work

Most general-purpose billing systems are designed for simple subscription models: $10/month, fixed seat counts, predictable invoices. That's not what Lago does. Because the reality is, usage-based billing touches everything:

1. Metering: Capture raw usage events (API calls, compute time)
2. Transformation: Normalize and aggregate usage (e.g., per org per hour)
3. Pricing: Apply business logic (tiers, discounts, overages)
4. Invoicing: Generate itemized, compliant, timezone-aware invoices
5. Payments: Handle retries, chargebacks, and dunning
6. Visibility: Expose real-time dashboards to customers and support teams

And then we haven't even talked about dunning campaigns, entitlements, maintaining edge cases and grandfathering in customers when you ship new pricing. This is an infrastructure problem disguised as a billing one.

Best practices and common pitfalls

Usage-based billing touches every part of your product and company. You need to get it right.

Automate invoice generation and payment collection from the start. Manual processes don't scale. Give customers real-time access to their usage data: if they can't see what they're accumulating, you'll get support tickets before you get renewals. Use rules-based pricing workflows that engineering doesn't have to touch every time pricing changes. Connect billing to the rest of the stack (CRM, support, analytics) so billing data flows where decisions get made.

What goes wrong

Don't overengineer early. Start with simple metrics and add complexity as the pricing model demands it. Batch-only processing is a liability: when a customer asks why their bill is $4,000 this month, you can't answer in real time. Keep billing logic auditable and testable. It's product code, not a spreadsheet. Always test pricing changes against real data in a staging environment before deploying. And if customers can't see and trust their usage, the billing system is doing damage regardless of how accurate it is technically.

Why we built Lago

We built Lago because we saw engineers wasting weeks on Stripe workarounds and brittle internal tools. We saw finance teams copy-pasting CSVs into spreadsheets to calculate overages.

We saw customers getting overbilled (or worse, underbilled) because no one trusted the data pipeline.

So we built an open-source billing engine that treats usage-based billing like what it is: a first-class product system. Real-time metering. Complex pricing models. API-first integrations. Self-hosted for companies with data residency or security requirements. Open-source so engineering can inspect and extend the billing logic instead of treating it as a black box.

We're building for engineers who care about billing. Because billing is product.

Built by Lago, open-source billing infrastructure for usage-based and hybrid pricing. Explore the docs or book a demo.