Pricing Teardown

    Usage-Based Pricing, Packaged for PLG

    The companies winning at PLG aren't using pure usage-based pricing — they're using hybrids. Here's the teardown, the playbook, and the mistakes to avoid.

    TL;DR

    • Pure usage-based pricing rarely works for PLG — it's too unpredictable for both sides.
    • The winners (Snowflake, Vercel, HubSpot, Datadog) use a platform fee + metered usage hybrid.
    • The right value metric grows when your customer grows — never charge for something they want less of.
    • Make the meter visible in-product. Surprise bills are the #1 reason usage-based PLG churns.

    The Four Pricing Models, Compared

    ModelExampleBest ForUpsideRisk
    Pure SubscriptionBasecampPredictable workflows, flat-rate buyersSimple to forecast, easy to sellHeavy users subsidized by light users; expansion is hard
    Per-SeatSlack, NotionTeam collaboration toolsScales with team adoption; clean unit economicsCustomers cap seats to control cost; punishes virality
    Pure Usage-BasedTwilio, AWS, OpenAI APIInfrastructure, API, and consumption productsRevenue tracks value delivered; near-zero friction to startUnpredictable bills erode trust; revenue is lumpy
    Hybrid (Platform + Usage)Snowflake, HubSpot, DatadogPLG companies moving upmarketFloor revenue + expansion upside; lands and expands cleanlyPricing page complexity; requires strong in-product metering

    Four Teardowns from Companies Getting It Right

    Snowflake

    Compute credits, decoupled from storage

    What they do

    Customers pre-purchase credits and burn them by the second on compute, billed separately from cheap storage.

    Why it works

    Aligns price with the exact moment value is created (running a query). Storage stays cheap so customers hoard data — which fuels more queries.

    The lesson

    Decouple your value metrics. Charge for the expensive, value-creating action — not the cheap thing that drives it.

    OpenAI

    Per-token pricing with model tiers

    What they do

    Pay per 1M input/output tokens. Cheaper models for volume, premium models for hard tasks.

    Why it works

    Token usage is a near-perfect proxy for value. Tiered models let the same customer self-segment between cost and quality on a per-call basis.

    The lesson

    Let customers trade off cost vs. quality inside your product. They'll spend more in aggregate than a single flat rate would capture.

    Vercel

    Free tier + metered overages on a platform fee

    What they do

    Generous free hobby tier, $20/user Pro with included usage, then metered bandwidth, build minutes, and function invocations.

    Why it works

    The free tier is the funnel. The platform fee anchors revenue. Metered overages capture upside without forcing a sales call.

    The lesson

    Use a platform fee as your floor and usage as your ceiling. Don't rely on pure usage — you need a recurring base to forecast.

    HubSpot

    Tiered platform with contact-based pricing

    What they do

    Free CRM, paid Hubs (Marketing, Sales, Service), priced by marketing contacts.

    Why it works

    Contacts grow as the customer succeeds, so price grows with value. Free CRM seeds the entire ecosystem.

    The lesson

    Pick a value metric that grows when your customer grows. If your metric goes up when they win, you've built an expansion engine.

    The 6-Step Playbook for Packaging Usage-Based Pricing

    1. Pick a value metric, not a cost metric

    The right metric is something the customer wants more of (API calls, contacts, rows synced, messages sent). The wrong metric is something they want less of (storage, seats they're forced to buy). If usage going up feels punishing, you picked wrong.

    2. Always pair usage with a platform fee

    Pure usage-based pricing is a forecasting nightmare. A monthly platform fee gives you predictable ARR, gives the customer a clear commitment, and gives sales something to land. Usage becomes the expansion lever.

    3. Make the meter visible inside the product

    Show usage, projected bill, and pacing in-app. Surprise invoices kill trust faster than any feature gap. AWS lost a decade of goodwill to this; don't repeat their mistake.

    4. Free tier ≠ freemium — design the gap

    Your free tier should let users hit the aha moment and then naturally bump into a usage limit that proves value. If they can run a whole business on free, the tier is too generous. If they can't see value, it's too tight.

    5. Cap exposure for new customers

    Offer spend caps, alerts, and soft limits during the first 90 days. Reduces churn from sticker shock and turns finance from a blocker into a champion.

    6. Re-price annually based on data

    Pull cohort data on usage distribution every quarter. Most teams under-price the top 10% by 2-5x. Move the price ladder, don't just add SKUs.

    Anti-Patterns to Avoid

    • Charging for storage as your value metric. Storage is cheap and feels punishing. Charge for what you do with the data.
    • Hiding the meter. If a customer can't see what they're spending in real time, your churn rate is a ticking clock.
    • No commitment floor. Pure pay-as-you-go gives you no ARR, no forecastability, and no leverage in renewals.
    • Pricing the customer's mistakes. If a customer accidentally racks up $50K overnight, your support cost just exceeded the revenue.

    FAQ

    Is usage-based pricing better than per-seat for PLG?

    For products where value scales with consumption (APIs, data, AI, infrastructure), yes. For collaboration tools where value scales with people in a room, per-seat is still cleaner. Most PLG companies above $10M ARR use a hybrid: platform fee plus usage.

    What is the best value metric for SaaS pricing?

    A good value metric is easy to understand, scales with customer success, and is something the customer actively wants more of. Examples: API calls, contacts, GB processed, messages sent, transactions. Avoid metrics customers want to minimize like seats or storage.

    How do I move from per-seat to usage-based pricing without churning customers?

    Run it as a hybrid first. Keep existing seat pricing as the floor, add a usage component for new value. Grandfather existing customers for 12 months, migrate new logos to the new model, then converge over the next renewal cycle.

    How much should a platform fee be in a hybrid model?

    Enough to commit the customer but low enough not to block self-serve. For PLG SMB, $20-100/month per workspace is typical. For mid-market, $500-2,500/month. The platform fee should cover your cost to serve plus a margin — usage carries the upside.

    Need help repricing your PLG product?

    I help B2B SaaS teams move from per-seat to hybrid usage-based pricing without churning their base.