Designing Usage-Based Pricing with Data

The usage metric most teams choose for consumption pricing is the one that is easiest to count. API calls. Events processed. Records created. These metrics are easy to instrument, easy to calculate, and easy to bill against.

They are also frequently wrong.

The metric you choose should not be determined by your engineering team's telemetry convenience. It should be determined by what predicts customer value. Getting from instrumentation convenience to value alignment requires data, not instinct.

The True Bill

Choosing a usage metric without data creates a specific type of commercial risk that compounds over time: you build enterprise contracts, sales compensation structures, and customer success metrics around a proxy for value that may not hold.

When the metric diverges from value, you see it first in expansion resistance. Customers who are using the product heavily but not achieving proportional outcomes push back on usage-based increases. Customers who achieve high outcomes at low usage feel overcharged on a per-unit basis when they scale. In both cases, the contract structure that should create automatic expansion instead creates negotiation friction.

Over 24 months, a misaligned usage metric in a $20M annual recurring revenue (ARR) usage-based business can suppress net revenue retention (NRR) by 8 to 15 percentage points relative to a well-aligned model. That is $1.6M to $3M in annual revenue difference, caused by a metric selection decision made at a time when the team did not yet have the data to make it well.

Execution

Step 1: Pull your NRR cohort analysis before touching your pricing model.

Segment your customers into NRR quintiles. For each quintile, pull the average values of every measurable product usage metric over the past 12 months. The metric that shows the steepest gradient from lowest NRR to highest NRR is the one most correlated with customer value. That metric is your leading candidate for a billing variable.

Step 2: Run the shadow billing analysis before committing to any metric.

Take your top 3 candidate metrics and simulate what each customer would have been charged over the last 12 months if you had billed against each metric. For each simulation, calculate: total revenue impact, revenue distribution by customer segment, and the implied price per unit for each cohort. Present these three models side by side. The model that produces the most intuitive alignment between customer size, value, and charge is the one to test first.

Step 3: Validate with 5 customer conversations before launch.

Take 5 accounts across different NRR cohorts. Show each account their simulated invoice under the proposed metric. Do not explain the metric first. Ask them what the invoice means to them and whether the charge feels proportional to the value they receive. The accounts in your top NRR cohort should find the simulated charge reasonable or low. The accounts in your bottom NRR cohort should find it somewhat high. If you see the reverse pattern, your metric selection is inverted.

Where It Unravels

A $16M ARR collaboration platform chose "active users per month" as their billing metric. They chose it because it was already tracked in their billing system and was a common metric in their category.

Shadow billing analysis, which they ran 6 months after launch rather than before, showed that their highest-NRR accounts had relatively low active user counts because they had efficient, senior teams who accomplished more with fewer people. Their lowest-NRR accounts had high active user counts because they had large, less efficient teams who churned due to low ROI.

The metric penalized efficiency and rewarded inefficiency. The company switched to "outcomes completed" within 10 months. The retrospective analysis showed that had they run the NRR cohort analysis before launch, they would have identified this problem in week 2.

Move This Week

Run the NRR cohort analysis for your top 3 candidate usage metrics. You need billing data, NRR by account, and usage logs. The analysis can be completed in a spreadsheet in a half-day if the data is available. If the data is not available, that is your first problem to solve.

Take the FintastIQ Pricing Diagnostic to see where your usage metric selection stacks up against similar-stage SaaS companies.

For related reading, see the diagnostic checklist for usage-based pricing in 90 days and first principles of usage-based pricing for SaaS.