Pricing teardowns
25 real products, the mechanics their pricing pages document, and an editable remix you can load into the simulator. Checked November 17, 2026.
25 of 25 teardowns
Developer tools · Hybrid
Cursor
Unit: metered model actions
Keep included usage and paid model spend legible: when the model a user picks changes how fast an allowance drains, show the drain rate next to the picker, not in a billing FAQ.
Read the Cursor teardownDeveloper tools · Hybrid
Lovable
Unit: build credits
When one product spends from several balances, label which balance each feature consumes at the moment it consumes it.
Read the Lovable teardownSupport · Per outcome
Intercom Fin
Unit: support attempts
Outcome pricing only charges for successes, but every attempt still costs you inference. Model the failed attempts before you pick an outcome price.
Read the Intercom Fin teardownInfrastructure · Hybrid
Vercel
Unit: normalized compute units
An included dollar credit is easy to sell but hard to reason about; show the credit next to the actual meter it is burning down.
Read the Vercel teardownCollaboration · Flat / seat
Linear
Unit: issues created
A volume fence can trigger the upgrade while revenue stays seat-based. Pick a fence your active free teams hit only once they depend on the product.
Read the Linear teardownAI apps · Flat / seat
Perplexity
Unit: advanced searches
On a flat plan, surface model and mode limits at the decision point so power users meet the limit as a choice rather than a surprise.
Read the Perplexity teardownInfrastructure · Hybrid
Supabase
Unit: GB egress
Teach the difference between a quota (what is included) and a spend cap (whether you may go past it). Defaulting the cap on protects new customers from surprise bills.
Read the Supabase teardownDeveloper tools · Hybrid
Resend
Unit: emails sent
Round-up buckets matter for low-volume customers: one email over the line costs a full block. The remix uses a simple per-email equivalent instead.
Read the Resend teardownCreative · Hybrid
Midjourney
Unit: Fast GPU minutes
Price speed separately from a slower path: a cheaper queue lets heavy users keep working without draining your expensive capacity.
Read the Midjourney teardownAI apps · Usage only
OpenAI API
Unit: thousand-token bundles
Show the token mix behind any estimate. A thousand-token bundle is a convenient blended unit, but input, output and cache rates move it a lot.
Read the OpenAI API teardownAI apps · Usage only
Anthropic Claude API
Unit: thousand-token bundles
Model cache hit rates before quoting one effective cost per action; a long-prompt band can change the rate entirely.
Read the Anthropic Claude API teardownAI apps · Usage only
Google Gemini API
Unit: thousand-token bundles
A free fence can restrict both which models are available and how fast they run. Decide which of the two actually protects your costs.
Read the Google Gemini API teardownDeveloper tools · Hybrid
Replit
Unit: publishing units
Track hosting burn separately from the agent workflow; a customer who publishes a busy app looks nothing like one who only builds.
Read the Replit teardownDeveloper tools · Hybrid
GitHub Copilot
Unit: AI credits
Explain both measures a buyer pays for: the seat and the workload. A team admin budgets seats; a heavy user drains credits.
Read the GitHub Copilot teardownCollaboration · Flat / seat
Notion
Unit: collaboration operations
Feature fences and scale fences can coexist with simple seat pricing; you do not need a meter to protect costs if the fence is placed well.
Read the Notion teardownCollaboration · Flat / seat
Figma
Unit: AI design actions
Make eligible roles and quotas visible before a team buys; a seat-type matrix is cheaper to explain up front than to dispute on an invoice.
Read the Figma teardownCollaboration · Flat / seat
Slack
Unit: messages modeled
Retention and access window are different product decisions. A visible-history fence converts teams without metering anything.
Read the Slack teardownInfrastructure · Hybrid
Cloudflare Workers
Unit: thousands of requests
A per-request limit is a cost cap that never needs a billing decision. Model requests and CPU separately in production.
Read the Cloudflare Workers teardownDeveloper tools · Hybrid
GitLab
Unit: hosted CI minutes
Distinguish the seat entitlement from infrastructure burn, and give heavy users a bring-your-own-capacity escape hatch.
Read the GitLab teardownInfrastructure · Hybrid
Sentry
Unit: thousands of error events
When you have several signals, keep each meter and its retention visible on its own line.
Read the Sentry teardownInfrastructure · Usage only
PostHog
Unit: thousands of analytics events
Explain drop versus bill behavior plainly. Customers forgive a dropped event far less often than they forgive a clear limit.
Read the PostHog teardownCreative · Hybrid
ElevenLabs
Unit: voice minutes
Show the conversion from credits to minutes. Buyers think in minutes of audio, not credits.
Read the ElevenLabs teardownCreative · Hybrid
Runway
Unit: generated video seconds
Price an actual generation workload before choosing a limit; credits per second differ enough between models to break a flat allowance.
Read the Runway teardownAutomation · Hybrid
Zapier
Unit: successful tasks
Define the event that consumes a unit in one sentence. “Successful action, not polling” is the kind of rule buyers can check themselves.
Read the Zapier teardownCollaboration · Flat / seat
Airtable
Unit: record operations
Model who gets billed before modeling the workspace's volume; billable roles move revenue more than record counts do.
Read the Airtable teardown