SaaS Pricing Strategies That Actually Work in 2025

We changed our pricing three times this year. Each time taught us something about what actually works.

Here’s the current state of SaaS pricing.

The Old Models Are Breaking

Per-Seat Pricing Problems

Traditional SaaS: Charge per user. Simple.

2025 reality: AI features have variable costs. A power user might cost you 10x what a light user costs. Per-seat doesn’t capture this.

Companies charging flat per-seat rates for AI features are either:

  • Losing money on power users
  • Overcharging light users
  • Both

Flat-Rate Problems

Unlimited usage for a flat fee sounds customer-friendly. But with AI costs, unlimited is unsustainable.

We offered “unlimited AI features” for one month. Our compute costs 4x’d. Never again.

What’s Actually Working

Usage-Based Pricing

Charge based on what customers actually use. Credits, API calls, tokens, whatever makes sense.

Pros:

  • Costs scale with revenue
  • Fair to light and heavy users
  • Incentivizes efficient product use

Cons:

  • Unpredictable for customers
  • Harder to sell than simple pricing
  • Requires good usage tracking

Who’s doing it well: Vercel, OpenAI, Stripe

Hybrid: Base + Usage

Flat monthly fee gets you access. Usage charges above certain thresholds.

Example: $99/month includes 1000 AI credits. Additional credits $0.10 each.

Pros:

  • Predictable base revenue
  • Scales with heavy users
  • Customers understand what they’re getting

Cons:

  • More complex than pure models
  • Requires explaining thresholds

This is what we settled on. Best of both worlds.

Outcome-Based Pricing

Charge based on results, not activity. Revenue share, per-transaction fees, success fees.

Pros:

  • Perfect alignment with customer value
  • Easy to justify price
  • Customers pay when they succeed

Cons:

  • Hard to track outcomes
  • Customers may not trust your tracking
  • Revenue is variable

Works for: Lead gen, marketplace features, sales tools. Harder for general productivity.

AI Feature Pricing Specifically

The hard question: How do you price AI features without losing money?

Option 1: Bake It In

Include AI in your base price. Absorb the cost. Use as differentiator.

Works if: AI is table stakes and you need it for competitive reasons. AI costs are small relative to subscription price.

Doesn’t work if: AI is the main value prop. Heavy users will destroy your margins.

Option 2: Separate Tier

AI features only in higher tiers. Forces upgrade for AI access.

Works if: You have clear tier differentiation. AI features are premium, not essential.

Doesn’t work if: Competitors include AI at all tiers. Tier structure becomes confusing.

Option 3: Credit System

Give credits monthly. Charge for overages.

Works if: Usage varies significantly between customers. You can predict reasonable credit allocations.

Doesn’t work if: Customers hate tracking credits. Your product requires constant AI use.

The Pricing Page Test

Show your pricing page to five potential customers. Ask:

  1. What do you think this costs?
  2. What do you get for that price?
  3. Is this good value?

If they can’t answer quickly, your pricing is too complex.

We failed this test twice before getting it right.

Price Anchoring Still Works

Psychology hasn’t changed:

  • Show three options, highlight the middle one
  • Expensive option makes middle seem reasonable
  • Annual discounts drive cash flow

What’s new: Showing AI costs separately (even if bundled) helps customers understand value.

“Includes $50/month of AI credits” is more compelling than “$50/month for features.”

The Discount Question

Should you discount?

Discount for:

  • Annual commitments (20-30% off)
  • Startups and nonprofits (if it gets you case studies)
  • Strategic customers

Don’t discount for:

  • “Just to close the deal”
  • Customers who don’t fit your ICP
  • Without getting something in return

Every discount trains customers to expect discounts.

Raising Prices

If you’re not uncomfortable with your prices, they’re too low.

How we raised prices:

  1. Grandfather existing customers for 6-12 months
  2. Announce early with clear reasoning
  3. Show what’s been added since they signed up
  4. Offer annual lock-in at old rates

Result: 3% churn. Less than expected. Should have raised earlier.

My Framework

For any SaaS product in 2025:

  1. Calculate your true costs per customer tier. Include AI and infrastructure.

  2. Design pricing around your most profitable customer type. Not the average.

  3. Build in usage components where costs vary. Don’t subsidize heavy users.

  4. Test with real customers before launching. Their reactions tell you everything.

  5. Plan to change it. First pricing is never right. Make it easy to adjust.

Price is the best marketing signal of value. Price like you believe in your product.