How 20 SaaS Companies Reduced Costs by Switching to Usage-Based Billing Software

Introduction

The most expensive line item in a SaaS billing stack is rarely the one labeled “billing software.” It’s the compounding cost that nobody budgets for: per-transaction fees charged as a percentage of revenue.

Stripe Billing charges 0.5–0.8% of billing volume. Chargebee adds metered billing at $599/month plus overages. Zuora scales to five-figure annual contracts. At $1M ARR these fees are invisible. At $10M ARR they become a budget line. At $50M+ ARR, they’re a material cost of revenue — often exceeding $300,000 per year for companies billing at scale.

The 20 companies profiled here made a deliberate infrastructure decision: they moved to usage-based billing software that gave them direct control over metering, pricing logic, and billing data — eliminating percentage-of-revenue fees and reducing billing overhead in the process. Some built their own infrastructure. Others adopted self-hosted platforms. All of them reached a point where the economics of third-party billing software didn’t hold up against the volume they were processing.

This article examines what drove each decision, what changed after the switch, and what the experience reveals about when it makes sense to take billing infrastructure in-house.

What Is Usage-Based Billing Software?

Usage-based billing software is infrastructure that meters customer consumption of a product or service and generates invoices based on actual usage — rather than fixed monthly fees. Also called consumption-based billing or metered billing, it replaces flat subscription pricing with variable charges tied to measurable units: API calls, data processed, compute hours, active users, transactions, or any custom metric.

A complete usage-based billing platform typically includes:

• Event ingestion and metering — capturing usage events in real time from application code or infrastructure
• Pricing engine — applying rate cards, tiers, and volume discounts to raw usage data
• Invoice generation — producing accurate invoices based on aggregated usage for each billing period
• Customer portal — giving customers visibility into their own usage and spend
• Revenue reporting — aggregating billing data for finance, forecasting, and audit purposes

The critical distinction between usage-based billing software and a payment processor like Stripe is scope. Stripe handles payment collection. Billing software handles the layer above it: determining what to charge, not how to collect it.

ABAXUS: Self-Hosted Usage-Based Billing Infrastructure

Idempotent event metering, versioned pricing, and automated invoicing — deployed inside your Kubernetes cluster. No per-transaction fees.

Book Architecture Review

Why Companies Switch: The Economics Behind the Decision

The Third-Party Billing Tax

SaaS billing platforms charge in two ways: flat subscription fees and a percentage of billing volume. For most companies at early stage, neither feels significant. The math changes at scale.

A company processing $5M/month in customer invoices through a billing platform charging 0.7% of volume pays $35,000/month — $420,000/year — in billing software fees alone. That’s before headcount to manage billing disputes, engineering time to work around platform limitations, or the cost of migrating data when contract terms change.

OpenView Partners’ research consistently shows that usage-based companies achieve 120% net revenue retention on average versus 105% for traditional SaaS — but that retention advantage erodes if billing overhead consumes the margin. Getting billing infrastructure right is both a revenue and a cost problem.

The Data Control Problem

Third-party billing platforms own your usage data. Switching vendors means migrating billing history, customer invoices, and usage records — a project that typically takes 6–18 months and involves significant engineering effort. Companies that reach scale often discover they’ve traded billing simplicity for vendor lock-in at the data layer.

Self-hosted or on-premises billing infrastructure keeps usage data inside your own systems, making audits, forecasting, and future migrations substantially easier.

The Customization Ceiling

SaaS billing platforms support common pricing models well: per-seat, per-unit, tiered. They struggle with complex, product-specific pricing logic: multi-dimensional usage rates, hierarchical account structures, custom commitment pricing for enterprise deals, or real-time budget alerts. Companies building differentiated pricing as a product feature — not just as an operational necessity — typically hit the customization ceiling of third-party platforms within 18–24 months of growth.

Photo by Jakub Żerdzicki on Unsplash

20 Companies That Took Billing Infrastructure In-House

Cloud Infrastructure Platforms (6 Companies)

Amazon Web Services built its own billing infrastructure from the beginning of the cloud era. AWS measures dozens of usage dimensions — compute hours, data transfer, API requests, storage capacity — and generates invoices for millions of customers monthly. The scale made third-party billing software impossible: no platform in 2006 could handle the event volume or the pricing complexity AWS required. Their pay-as-you-go model grew to support $90B+ in annual revenue, with billing infrastructure that processes billions of usage events daily.

Snowflake separated compute from storage and built usage metering around “credits” — a proprietary unit that abstracts query compute cost. Rather than charge per query (unpredictable for customers) or per seat (misaligned with value), Snowflake engineered a consumption model that let customers control their spend while giving Snowflake a revenue stream that expanded naturally with customer workload growth. Their consumption model contributed to $2.8B in revenue (FY2024) and net revenue retention consistently above 130%.

DigitalOcean built transparent, predictable billing around compute, storage, and bandwidth consumption. Rather than abstracting usage behind complex pricing tiers, DigitalOcean made billing legibility a product feature — customers always knew what they’d pay. Their billing stack scales across millions of customers without per-transaction billing fees, contributing to $693M revenue in 2023.

Heroku pioneered the “dyno” as a billing unit — a standardized compute container that made billing predictable and scalable simultaneously. Developers could provision exactly the compute they needed, and Heroku billed per dyno-hour. This model handled usage spikes without manual upgrades and gave Heroku a billing primitive that matched how developers think about application infrastructure.

Netlify built bandwidth and build-minute metering directly into its platform. Free tiers enabled developer adoption; usage-based pricing above free limits converted power users into paying customers without a sales motion. Netlify’s billing model is a textbook example of product-led growth enabled by consumption pricing.

Vercel meters function execution, edge requests, and bandwidth consumption. Their billing infrastructure supports a freemium motion where individual developers adopt the product and usage-based billing converts teams and enterprises automatically as workloads scale. Vercel’s model generated a $1B+ valuation with a substantially smaller go-to-market team than comparably-sized SaaS companies.

API and Developer Platform Companies (6 Companies)

Twilio standardized per-API-call billing for communication infrastructure. Each SMS, voice call, or video session generates a discrete billing event. Twilio’s billing system ingests millions of these events daily, applies carrier rates, marks them up, and generates invoices across hundreds of thousands of customers. Their pricing model — transparent, predictable, aligned with developer usage — fueled growth from $2M ARR to $4.1B in revenue (2023) without requiring enterprise sales cycles for most customer segments.

Stripe charges per transaction at 2.9% + $0.30. That model is simple by design — but the billing infrastructure behind it is complex. Stripe’s platform accounts for disputes, refunds, partial captures, currency conversion, and tax compliance across 135+ currencies and every major payment method. Stripe’s billing stack is effectively a purpose-built financial ledger that processes hundreds of billions in payment volume annually. At that scale, a third-party billing tool isn’t an option — it’s the product.

SendGrid built email volume billing that scaled from startup developers sending 10,000 emails/month to enterprise senders processing billions. Their billing infrastructure handled volume tiers, overage pricing, and dedicated IP allocation — all metered automatically. The billing model contributed to SendGrid’s acquisition by Twilio for $3B in 2018.

Segment charged based on Monthly Tracked Users (MTUs) — a metric that aligned precisely with the value customers received (more tracked users = more insight). The MTU billing model required custom metering infrastructure that counted unique users across event streams rather than simply counting events. This nuanced metering made Segment’s pricing defensible and their billing infrastructure proprietary — a combination that supported their $3.2B acquisition by Twilio in 2020.

Algolia bills per search operation, with pricing that scales across billions of queries monthly. Their billing infrastructure captures search events in real time, aggregates them by customer, applies pricing tiers, and generates invoices — all without percentage-of-revenue fees to a third-party platform. Algolia’s metered pricing model supports $200M+ in ARR with a billing stack entirely under their own control.

PlanetScale built database-native billing around storage, rows read, and rows written — granular metrics that let customers understand exactly what drives their bill. Their billing model required purpose-built usage metering that integrated directly with their MySQL-compatible database layer, not a general-purpose billing platform.

Data and Analytics Platforms (4 Companies)

Databricks bills on Data Processing Units (DPUs) — a compute metric that abstracts cluster configuration from billing. Rather than charging per VM or per cluster-hour (which exposes Databricks to infrastructure cost fluctuations), DPU billing gives Databricks pricing leverage while giving customers a stable unit to budget against. Databricks reached $1.6B in revenue (FY2024) with a billing model that required custom metering infrastructure integrated directly with Apache Spark job execution.

Elastic transitioned from open-source to a consumption model charging based on data ingestion volume and deployment size. The switch required rebuilding billing infrastructure around usage events rather than seat counts — a multi-year engineering project that repositioned Elastic from a commoditizing open-source product to a managed service with predictable revenue expansion. Elastic reports $1.3B in annual revenue (FY2024).

New Relic migrated from per-host subscription pricing to a data-ingest model in 2021. The migration required new metering infrastructure, new pricing logic, and a complete renegotiation of how customers thought about their spend. New Relic’s transition demonstrates both the complexity and the payoff of moving billing models at scale: the new data-ingest model expanded their addressable market to customers who previously couldn’t budget for per-host pricing.

Datadog meters hosts, custom metrics, log volume, and APM traces — four separate usage dimensions that often scale independently for each customer. Their billing infrastructure aggregates across these dimensions and applies volume discounts at the account level. This multi-dimensional metering requires custom billing infrastructure; no off-the-shelf billing platform handles the intersection of four independent usage dimensions with account-level aggregation and volume discounts.

E-Commerce and Payments Infrastructure (4 Companies)

Shopify built revenue-share billing for Shopify Plus merchants using their payment infrastructure. Merchants pay a percentage of gross merchandise value (GMV) above a threshold — a model that requires Shopify’s billing system to track transaction volume, apply thresholds, and calculate variable fees per merchant monthly. This billing model contributed to $7.1B in total revenue (2023), with merchant solutions (primarily payment processing and billing fees) representing the fastest-growing segment.

BigCommerce offers GMV-based billing tiers where merchants upgrade plans as their online revenue grows. The model requires BigCommerce to monitor merchant sales volume and trigger plan changes automatically — a billing operation that requires their own usage tracking rather than relying on a third-party billing tool.

Recurly — notable as a billing platform itself — meters subscription and usage events for thousands of SaaS companies. The irony is instructive: Recurly built its own billing infrastructure to support the billing needs of others. The platform handles dunning, proration, billing disputes, and revenue recognition — each of which requires proprietary logic that a general-purpose billing tool couldn’t provide.

MongoDB Atlas charges on a combination of storage, compute tier, and data transfer. Their billing infrastructure aggregates usage across clusters in multiple cloud regions, applies regional pricing, and generates consolidated invoices for enterprise customers with multi-cluster deployments. Atlas reached $1.7B in annualized revenue with billing infrastructure that handles usage complexity no third-party platform could support out of the box.

The Cost Reduction Pattern: What These Companies Have in Common

Across these 20 companies, three consistent cost-reduction themes emerge:

1. Eliminating percentage-of-revenue billing fees. Every company that owns its own billing infrastructure pays a fixed cost to operate that infrastructure rather than a percentage of revenue processed. At scale, the difference is material. A company processing $20M/month in customer invoices through a third-party billing platform paying 0.7% spends $1.68M/year in billing fees. The equivalent self-hosted infrastructure costs a fraction of that.

2. Reducing billing accuracy failures. Third-party billing platforms introduce integration complexity — events must travel from your application through an API to an external system before appearing on an invoice. Each integration point is a potential failure mode. Companies that own their billing infrastructure report significantly lower billing dispute rates because usage data flows directly from their own systems rather than through third-party APIs.

3. Removing the customization constraint on pricing strategy. Every company in this list has pricing that no off-the-shelf billing platform could implement without heavy customization: DPUs (Databricks), MTUs (Segment), dyno-hours (Heroku), credit-based compute (Snowflake). Companies that own billing infrastructure can implement any pricing model their product strategy requires — rather than constraining pricing to what their billing vendor supports.

When It Makes Sense to Take Billing In-House

Not every company needs to build billing infrastructure from scratch. The decision depends on three factors:

Volume thresholds. At under $2M ARR, third-party billing platforms are usually the right choice — the engineering cost of custom infrastructure exceeds the fee savings. Above $10M ARR processing significant transaction volume, the economics shift. Above $50M ARR, custom billing infrastructure is almost always justified on cost grounds alone.

Pricing complexity. If your pricing model fits cleanly into per-seat or simple per-unit tiers, standard billing platforms handle it well. If your pricing involves multi-dimensional usage, real-time budget caps, hierarchical account structures, or custom commitment pricing, you’ll hit the platform ceiling faster than you expect.

Data control requirements. Companies in regulated industries — financial services, healthcare, government — often can’t send billing events to third-party SaaS platforms. Self-hosted billing infrastructure keeps usage data inside your own security perimeter.

Running into billing infrastructure limits?

ABAXUS deploys inside your Kubernetes cluster — giving you the metering accuracy and pricing flexibility of purpose-built infrastructure without the engineering cost of building it yourself. See pricing starting at $4,800/yr.

See Pricing

Conclusion

The 20 companies profiled here didn’t switch to usage-based billing software because it was trendy. They made an infrastructure decision: the economics of third-party billing platforms didn’t hold up against the volume they were processing, the pricing complexity their product required, or the data control their customers demanded.

AWS, Snowflake, Twilio, and Databricks built their own billing infrastructure because no vendor could support their requirements. Recurly, Stripe, and MongoDB Atlas built proprietary billing as a product feature that differentiated their offering. What unites all 20 is that billing stopped being a vendor relationship and became an engineering asset.

For companies reaching the same inflection point — where third-party billing fees are becoming a material cost, where pricing customization is constrained by platform limits, or where usage data needs to stay inside your own infrastructure — the options have historically been: build it yourself (expensive, slow) or accept the constraints of SaaS billing vendors (costly at scale).

ABAXUS provides a third path: self-hosted usage-based billing infrastructure that deploys inside your Kubernetes cluster. You get idempotent event metering, versioned pricing logic, automated invoicing, and complete data ownership — without percentage-of-revenue fees and without the 18-month engineering project of building billing infrastructure from scratch. Annual licenses start at $4,800/yr.

If your billing architecture is becoming a constraint on your pricing strategy or your unit economics, book an architecture review to map your current event sources, identify data contract gaps, and see what fixing the architecture looks like.