How to price an AI product: value, tiers, packaging

Time: about 35 min reading + 270 min practice (steps 1–3; steps 4–5 come after launch)

The dollar amounts, percentages and pricing ladders in this lesson are made-up examples that explain the method. They aren't market statistics, and they aren't a forecast of what you'll earn. Work out your own numbers for your own market and your own customers. For the beginner version, see the lesson How to set your price. This lesson is about pricing a product with a subscription: your own AI service, or a service with a monthly fee. If you only sell one-off projects for now, read the theory and come back to the practice once you have a product.

The gist

Pricing isn't just "what it costs." It's a signal of value, a filter for customers and your source of revenue. Three jobs in one number.

Many founders are afraid of pricing. "I'll charge $9 so at least someone buys," and then they can't raise it for two years. Meanwhile, competitors charge $99 for the same features and bring in noticeably more money with the same number of customers. They often get better customers, too, and higher retention (more customers stay).

Price isn't the last step in building a product. It's the first signal a customer sees after the description. $9 says "this is a toy." $999 says "this is a serious tool." What's inside may be identical, but the customer's perception, expectations and behavior are different.

For AI products the problem is sharper: the value can be tens or even hundreds of times the cost. The API (pay-per-use access to an AI model) costs $0.10, and the value to the customer is $50. Charge $0.20 (cost plus margin) and you leave $49.80 on the table. A competitor will take it.

In this lesson we'll cover how to price an AI product the right way: the value-based approach, a 3-tier structure, psychological triggers, and a 30-day plan for testing your price.

🎨 Picture this: a restaurant. The chef doesn't set the price as "what the ingredients and the gas cost." The price is "what a guest is willing to pay for this evening." A steak costs $80 not because the beef is more expensive, but because of the atmosphere, the name and the expectations. The same steak at a roadside diner costs $15. Similar product, different value, different price. An AI product works the same way.

🎯 The big shift: from cost thinking to value thinking

The most expensive pricing mistake is to calculate "bottom up": your cost → margin → price. That works for commodities (gas, rice, hardware bolts). For an AI product, it usually backfires.

The right approach is "top down": value to the customer → a percentage of that value → price.

Cost thinking (bad for AI):

• "The API costs $0.10, I'll add a margin → $0.20"
• Result: a race to the bottom on price

Value thinking (right for AI):

• "The customer saves $5,000 a month → a fair price is 10-20%, so $500-1,000 a month"
• Result: a healthy margin and premium positioning

🎨 Picture this: Uber vs. a metered cab. The meter counts miles (cost). Uber looks at what you're willing to pay for a fast ride at rush hour (value). That's how Uber can charge more for the same route at the busiest times.

Key concepts

• Value-based pricing: the price is a share of the value the customer gets, not a function of your cost
• Tier structure: 3 levels (Starter / Pro / Business) that capture different customer segments
• Rule of 3x: each tier costs about 3 times the previous one ($29 → $99 → $299)
• Anchor pricing: show the expensive option first so the middle one feels affordable
• Decoy tier: a "bad" option placed between good ones that nudges people toward the one you want them to pick
• Grandfathering: existing customers keep their old price when you raise prices (protects retention)
• Usage-based vs. flat: pay per use vs. a fixed subscription, plus hybrids of the two
• Annual discount: a discount for paying for a year up front (often 15-20%); it improves cash flow and retention
• Freemium vs. free trial: two different ways in. Freemium is a free plan with no time limit next to the paid plans; a free trial is the paid version for a limited time
• Willingness to pay (WTP): the most a customer is willing to pay; your price should sit below it

Theory

3 basic pricing models, and why two of them are risky for AI

Any pricing approach comes down to one of three philosophies: count up from your costs (cost-based), look at your neighbors (competitive), or measure what the customer gets (value-based). For an AI product, the right answer is almost always the third. But it's important to understand why the first two work poorly.

Model A: Cost-based (bad for AI)

The approach: "The API costs $0.10, I'll add a 100% margin → I sell it for $0.20."

The problem for AI: the value can be tens or even hundreds of times the cost. If an AI tool saves a customer 10 hours of work ($500-1,000 of value) and you price by cost at $5, that customer would happily have paid $100. You leave $95 on the table with every sale.

When cost-based pricing makes sense:

• Commodity products (cloud hosting, raw API resale)
• Low differentiation (your product is the same as 10 others)
• A volume play (you win on volume, not on margin)

For AI products: almost never. Differentiation in AI is huge: the user experience, the prompts, the integrations, the domain expertise. It's not a commodity.

Model B: Competitive (risky for AI)

The approach: "ChatGPT Plus is $20/month (as of October 2026), so I'll charge $20/month too."

The problem: copying ignores that the value you deliver is different. ChatGPT is general-purpose. Your product may be 10 times more valuable for a narrow niche. Or 10 times less valuable for general use.

Pricing based on competitors is a sign that you don't have your own understanding of your value.

When competitive pricing makes sense:

• Entering a market late with something clearly better or cheaper
• A direct alternative (your product does the same thing, just better on one dimension)
• Deliberate positioning ("we're 50% cheaper than X")

For AI products: rarely. Many AI products target a new use case or a specific niche, where there's no direct comparison.

Model C: Value-based (right for AI)

The approach: measure the value to the customer → charge 10-20% of that value.

The logic:

• The customer gets $5,000 a month in value → they're willing to pay $500-1,000 a month
• It's an "easy yes": a clear ROI (return on investment), and each month's fee pays for itself within the first week
• The margin is high (cost $50-100, revenue $500-1,000)
• The customer is happy (they save $4,000-4,500 net)

This is a common approach to AI pricing in 2026. Many AI SaaS products (software sold as a subscription) work this way: the price is tied to the value for the customer.

🎨 Picture this: a family doctor vs. a surgeon. A family doctor charges, say, $150 for an office visit (the going market rate). A surgeon charges $50,000 for an operation that saves a life, because that value can't be compared with a standard rate. AI is often the "surgeon": it solves an expensive problem fast. Your pricing should reflect that.

Value-based pricing in 4 steps

The theory is clear. Now, how to actually calculate it.

Step 1: Quantify the customer's value

Value comes in four types. For any given AI product, usually 1 or 2 of them apply.

Time saved (the most common for AI):

• How many hours a month does the customer save thanks to your product?
• Multiply by the customer's hourly rate (or their employees')
• Result: $ value per month from time savings

Example: AI writes 50 emails instead of a person. Each email is 15 minutes of manual work. 50 × 15 minutes = 12.5 hours a month. A junior marketer's hourly rate is $40. Value = 12.5 × $40 = $500 a month.

Revenue uplift:

• The % increase in revenue your product delivers
• × the customer's current revenue
• Result: $ value per month from the revenue boost

Example: an AI sales tool lifts sales by 5%. The customer does $20,000 a month in sales. Uplift = $1,000 a month. Value = $1,000 a month.

Cost reduction:

• The % reduction in existing costs
• × the customer's current costs
• Result: $ value per month in savings

Example: AI customer support cuts support costs by 30%. The customer spends $10,000 a month on support. Reduction = $3,000 a month. Value = $3,000 a month.

Quality improvement (harder to quantify, but real):

• Lower risk (legal violations, churn, meaning customers canceling, and so on)
• A better customer experience
• Brand reputation

This type is harder to put into numbers, but don't ignore it: for enterprise customers it's often the deciding factor.

Step 2: Position yourself in the value range

Once you know the value, you pick a percentage of it as your price. (The previous lesson used 10-30% of the yearly value: that was the one-time price of a project. Here we're pricing a subscription: the percentage is taken from the monthly value and paid every month.)

Example: the value is $3,000 a month.

• Conservative: $300/month
• Standard: $450/month
• Premium: $600/month

For a new product with no proof yet, start conservative ($300). Once you've built up case studies, move up to standard or premium.

Don't charge more than a quarter of the value (a rule of thumb, not a law). Beyond that, the customer starts to feel the payback takes too long.

Don't charge too little (under 5% of the value) either. It signals the product isn't serious, and your margin won't cover the cost of support.

Step 3: Tier structure (always 3 tiers)

A single tier usually leaves money on the table. Common advice for SaaS pricing: at least 3 tiers.

The logic (a made-up split; yours will differ):

• a smaller share of customers will pick Starter (they need little hand-holding and are price-sensitive)
• most will pick Pro (the main offering, a balance of value and price)
• a small share will pick Business (an enterprise feel, high willingness to pay)

Without a top tier, you lose some revenue at the high end. Without a bottom tier, you lose your entry point to the mass market.

The standard structure:

Some companies add a 4th tier, Enterprise (custom pricing, sold through a sales team). But that sits on top of the 3 base tiers.

Step 4: Test in the market

Pricing isn't fixed forever. It's a system you keep iterating on.

Launch playbook:

1. Set your starting price (based on the value calculation)
2. Track for 30-60 days: conversion rate (the share of people who buy), churn rate, revenue per customer, support load
3. Iterate: if conversion is very low, lower the price or improve the offer. If every customer picks the top tier, raise your prices.
4. Repeat every 6-12 months

Signs your price is too low:

• Every customer says "that's cheap"
• Zero churn (even poor-fit customers don't leave)
• Suspiciously high conversion (too easy a yes)
• The product comes across as low-end in reviews

Signs your price is too high:

• Very low conversion
• Very few trial users go on to pay
• Sales calls drag on over price negotiation
• Churn is concentrated in the first 30 days

Tier design framework

Designing tiers is a craft of its own. Bad design means customers can't tell which one to pick, so they choose the cheapest one or don't buy at all.

Bad tier design: anti-patterns

Anti-pattern 1: tiers too close together

• Starter $9 / Pro $19 / Business $29
• The problem: everyone picks Pro (a small step up), Business isn't justified, and Starter makes people wonder "what am I missing?"

Anti-pattern 2: tiers too far apart

• Free / $9 / $99
• The problem: the gap between $9 and $99 scares people off, and the customer can't tell what's inside.

Anti-pattern 3: feature scatter

• Random features in each tier, with no logical progression
• The problem: the customer doesn't understand what they're getting.

Anti-pattern 4: too many tiers

• 6 tiers with different options
• The problem: decision paralysis. The customer leaves "to think about it."

Good tier design: the rule of 3x

Each tier costs about 3 times the previous one. That's enough for the customer to see a clear difference without being scared off.

Example (B2B SaaS):

CSM is a customer success manager, a person assigned to the account. SLA is a service level agreement, a written guarantee of uptime and response times.

Notice:

• $29 → $99 → $299: the rule of 3x at work
• Features are added progressively (not at random)
• Limits grow tenfold with each tier
• Enterprise is open-ended at the top and sold through a sales team

The decoy tier: a psychological trick

Sometimes a tier is added to nudge people toward the target one.

Example: you want most customers to pick Pro at $99. You set up:

• Starter $29 (basic)
• Pro $99 (the target)
• Business $99/month billed yearly (the same as Pro, only locked in for a year): the DECOY

The customer sees: "For the same $99 I can get Pro month to month, or the same thing locked in for a year." They pick Pro (month-to-month flexibility) or Business (if a yearly commitment suits them anyway).

A less cynical way to see it: a decoy simply helps the customer make a decision. Don't abuse it.

Pricing levers: what you can vary

Within value-based pricing and 3 tiers, there are levers that change exactly how you charge.

A hybrid usually beats a pure model:

• Pure per-usage pricing scares customers ("how much will this end up costing me?")
• Pure flat pricing doesn't scale with large users
• Hybrid: a base subscription + overage charges (fees for use beyond what's included) for heavy usage

🎨 Picture this: a cell phone plan. Paying for every megabyte means you're nervous every time you open a video (the meter's running). Unlimited means light users overpay. A hybrid (a set amount of data included, a charge for anything extra) gives you the best of both. AI pricing works the same way.

AI-specific pricing nuances

Pricing an AI product has pitfalls you don't run into with traditional SaaS.

The usage-based pricing trap

If you charge for every "request," customers are often afraid to make requests ("it'll get expensive"). They use it less → get less value → leave.

The fix: a hybrid. A base subscription (N requests a month included) + overage for those who go over. The customer knows their floor and isn't afraid to use the product, and you get predictable revenue.

The token transparency dilemma

Should you show "tokens" (the small chunks of text AI models are billed by) on your pricing page? It's a dilemma:

• Show them: a technical audience understands, a non-technical one gets lost
• Hide them: a black box, which breeds distrust
• Compromise: show "AI requests" or "credits" instead of tokens. Everyone understands those.

Example: "10,000 credits/month ≈ 500 long emails (10 credits each) + 200 code reviews (25 credits each)."

Cost variance (important for AI)

Some users can use 100 times more than others. With single-tier flat pricing, your heaviest users eat your margin.

A plan for outliers:

• A hard cap (after X usage, the service slows down or stops)
• Per-usage overage (gentler, but unpredictable for the customer)
• An upgrade prompt ("You're using a lot. Want to move to Pro?")

Many AI services use usage limits and communicate them clearly. That's fine as long as the limit is reasonable.

How to build your own pricing benchmarks

Want a sanity check on your prices? Build your own table from the public pricing pages of competitors in your niche. AI product prices change every few months, so there's no ready-made table of numbers here.

How to fill it in: take 5-10 direct competitors, open their pricing pages, and write down the prices of their three tiers and the date you checked. For an example of current prices for AI assistants and APIs, see What's current.

Use it as a guide, not a law. Your value may be higher or lower, and your pricing should reflect that.

Psychological triggers: what works in 2026

Pricing isn't only math. The psychology of perception has a noticeable effect on conversion.

Ending in 9 (charm pricing)

• $29 feels lower than $30 (the brain sees "twenty-something")
• Often works in B2C (selling to consumers)
• Probably matters less in B2B (selling to businesses), but doesn't hurt

Anchor (expensive option first)

• Show the Enterprise tier first on the page
• The middle tier then feels affordable
• Put a "Recommended" badge on your target tier; say "Most popular" only once it's true

Decoy (an asymmetric option)

• A "bad" tier between good options pushes people toward the target one
• Use it carefully, and don't overdo it

Bundle

• "Everything included" is psychologically easier than paying per feature
• The customer doesn't want to calculate "do I need this feature?"
• Simplicity wins on a pricing page

Money-back guarantee

• "30-day money-back guarantee" lowers the perceived risk
• It often helps conversion
• When the product is good, refund requests tend to be uncommon (offer the guarantee only if you're ready to give the money back)

Showing the annual discount

• "Pay $278 a year, save $70" (a $29/month plan at 20% off) lands harder than "$23/month" (the same yearly price divided by 12)
• Show the savings explicitly
• Default to monthly, highlight annual

Annual pricing strategy

An annual subscription is one of the strongest levers for cash flow and retention. But it takes care.

The standard discount: 15-20% off for annual.

The upside:

• Cash flow: you get roughly 10 × $X today instead of $X every month
• Retention: a psychological commitment, so churn is usually lower
• LTV: customer lifetime value (the total a customer pays you over time) goes up

The downside:

• Lock-in feels like a commitment (hurts trial conversion)
• Refunds get more complicated (a year paid up front means a partial refund process)
• Price changes get harder (annual customers wait for their term to end)

The compromise:

• Default: monthly
• Annual visible, with the savings spelled out ("Save $70/year")
• Let customers switch from monthly to annual at any time

Don't go annual-only. It cuts out part of your market: the customers who want to try the product without a commitment.

Price changes: how to raise prices

After 12-18 months, you'll want to raise prices. You'll have more features, more case studies, a stronger brand. That's normal.

Playbook:

1. Grandfather existing customers

• Existing customers keep their old price (forever, or for 2 years)
• This builds loyalty and prevents waves of cancellations
• The risk: they never upgrade. That's okay.

2. Announce it 30-60 days ahead

• Email + an in-app notice
• Explain the "why" (new features, more value)
• Say it plainly: the new price, the date it takes effect and how to cancel. Don't hide the increase behind vague wording
• In some places the law sets how and when you must give notice of a subscription price increase; check the rules where your customers live

3. Frame it as more value

• A new tier with extra features
• Old features are still available in the existing tier
• New sign-ups go on the new pricing

4. Test new pricing on new sign-ups first

• For 30-60 days, new sign-ups see the new prices
• Measure the impact on conversion and churn
• Adjust before the full rollout

5. Communicate value before price

• Start the email with "what's new" (3 features)
• Then state the new price and the date clearly. Don't bury it in fine print at the bottom
• The customer should see at a glance what changes and when

🎨 Picture this: a landlord who renovates. You don't tell your tenants "you're paying more now." You say "I redid the place: new kitchen, new bathroom, so here's the new rent." The way people take it changes completely.

Common pitfalls

7 mistakes many founders make, and what they cost.

❌ Pricing too low "to get started"

• Too low a price in B2B = no perceived value
• Customers may say it outright: "it's cheap, so it's probably not good"
• Raising it later is hard (see grandfathering)

❌ Pricing too high without proof

• Price = expectations. At $999/month, the customer expects enterprise-level support
• If you can't deliver, you get churn and bad reviews
• Raise prices as the product matures

❌ A free tier that gives away too much

• If the free tier solves 80% of the problem, there's no reason to upgrade
• Free should be a trailer (say, 5-10% of the capability) for the paid plans
• Notion and Figma have well-balanced free tiers

❌ Custom pricing for everyone

• It slows down the sales cycle a lot
• Self-serve becomes impossible
• Use custom pricing only for the top 5% (over $2K/month)

❌ Changing prices often

• Every 3 months = confusion + churn
• At most 1-2 changes a year
• Write down the "why" for transparency

❌ Annual without a monthly option

• It shrinks the pool of people who could buy
• New customers want to try it without a commitment
• Always offer both

❌ Not raising prices for 2 years

• Inflation alone opens a gap
• It looks cheap (the signal: low quality or a dying product)
• Review your prices at least once a year

A/B testing prices: handle with care

A/B testing prices isn't like A/B testing a user interface. The risk is high.

DON'T:

• Show different prices to different customers on the same page at the same time
• PR risk: someone posts a screenshot and it blows up on social media
• Damage to trust

DO this instead:

• Sequential A/B: 30 days at price X, 30 days at price Y, then compare
• Cohort A/B: people who signed up before a set date see the old price, and people who sign up after it see the new one. Be open about it ("the old price is available until such-and-such date"), and only if that's true
• Page-level A/B: different landing pages lead to different pricing pages (when the traffic sources are clearly different)

Measure several things together:

• Conversion alone is misleading (low price → high conversion → low revenue)
• Revenue per visitor = the main metric
• Plus: 30-day churn, NPS (a 0-10 "would you recommend us?" score), support load

Test one variable at a time:

• Change only the price (not the button color and the price at once)
• If you change several things, you won't know what worked

Where you are now: pricing by stage

Your pricing strategy depends on your stage.

Beginner (first product, under 100 paying customers):

• A simple 3-tier setup with an annual option
• Sample starting points: $29 / $99 / $299 (adjust them to your niche)
• Skip A/B testing (not enough data)
• Skip custom enterprise deals (they eat your time)
• Focus: get to product-market fit (a product the market clearly wants); don't optimize pricing yet

Intermediate (100-1,000 customers, an established product):

• Value-based pricing with a quantified $ value per month
• Usage tracking (by groups of customers who joined at the same time)
• A/B testing prices on new sign-ups
• Consider freemium if it's a mass-market play
• Annual + monthly options

Advanced (1,000+ customers, expanding):

• Dynamic enterprise pricing with a sales team
• Multi-product bundles (cross-selling)
• Custom contracts for the top 10%
• A quarterly pricing committee review
• Considered price increases (how much: decide from your data)

Practice

Step 1: Quantify your customer's value (90 minutes)

Grab a sheet of paper or open a Notion page. Answer 4 questions:

1. Which type of value dominates?

• Time saved
• Revenue uplift
• Cost reduction
• Quality improvement

Pick the top 1 or 2.

2. Put it in dollars:

Time saved formula:
Hours saved/month × Hourly rate = $value/month

Revenue uplift formula:
% increase × Current revenue = $value/month

Cost reduction formula:
% reduction × Current cost = $value/month

3. Make a conservative estimate:

Take the low estimate (not the optimistic one). Better to undersell the value in your pricing and overdeliver in reality.

4. Document it, with a source:

Not "well, probably $5,000 a month." Instead: "From interviews with 5 customers (or potential customers): average time saved 12 hours/month × $50/hour = $600/month."

Output: a "Customer value analysis" document with a specific $ value figure.

Step 2: Design a 3-tier structure (60 minutes)

Use this template:

# Pricing Tiers — [Product Name]

## Tier 1: Starter — $[X]/month
**Target:** Individual / small team trying it out
**Limit reasoning:** [why these particular limits]

Features:
- [Feature 1]
- [Feature 2]
- [Feature 3]

Limits:
- [N] users
- [N] AI requests/month
- [Storage / data / etc.]

Support: Email (48-hour response)

## Tier 2: Pro — $[3X]/month ← TARGET (the one you want most customers to pick)
**Target:** Growing business, main offering
**Limit reasoning:** [why]

Features:
- Everything in Starter +
- [Feature 4]
- [Feature 5]
- [Feature 6]

Limits:
- [N] users (5-10 times Starter)
- [N] AI requests/month (10 times Starter)
- Priority support

Support: Email + Slack (24-hour response)

## Tier 3: Business — $[9X]/month
**Target:** Large companies / heavy users
**Limit reasoning:** [why]

Features:
- Everything in Pro +
- [Feature 7]
- [Feature 8]
- Custom integrations
- Dedicated CSM

Limits:
- Unlimited users
- [N] AI requests/month (10 times Pro)

Support: Dedicated CSM + Slack channel

## Tier 4 (optional): Enterprise — Custom
**Target:** Top 5-10% of prospects
- Custom contract
- SLA
- Sales-led
- Min commitment: $[amount per year]

Sanity check:

• Does the rule of 3x hold? (Starter × 3 ≈ Pro, Pro × 3 ≈ Business)
• Is each tier a clear step up in value (not random features)?
• Do the limits grow 5-10 times between tiers?
• Is the target tier (Pro) clearly the best value?

Step 3: Build your pricing page (120 minutes)

A pricing page is a conversion page, not an information dump. These 7 elements are a must.

# Pricing Page Checklist

## 1. Hero section (the top of the page)
- [ ] One main headline (H1): "Pricing that scales with you" (or similar)
- [ ] Subheadline: a value statement (NOT a feature list)
- [ ] Monthly / Annual toggle (annual = 20% off)

## 2. 3 pricing cards
- [ ] Starter card
- [ ] Pro card (with a "Recommended" badge; use "Most Popular" only once it's true)
- [ ] Business card

Per card:
- [ ] Tier name
- [ ] Price (large)
- [ ] Short value statement (1 line)
- [ ] Feature list (5-7 items, no more)
- [ ] CTA button ("Start free trial" / "Contact sales")

## 3. Enterprise CTA
- [ ] "Need more? Contact sales" link
- [ ] Custom requirements ("100+ users, custom integration")

## 4. Trust signals
- [ ] "30-day money-back guarantee" (only if you're ready to honor it)
- [ ] Customer logos (3-5, only real customers, with their permission)
- [ ] Testimonials (1-2 on the pricing page, only real ones)

## 5. FAQ section (5-7 questions)
- [ ] Can I change plans?
- [ ] What payment methods do you accept?
- [ ] How does usage work?
- [ ] Can I cancel anytime?
- [ ] Is there a free trial?
- [ ] Do you offer a nonprofit / education discount?
- [ ] What's your refund policy?

## 6. Feature comparison table
- [ ] Detailed table: all features × all tiers
- [ ] Helps users compare specifically
- [ ] Checkmarks for clarity

## 7. Footer
- [ ] Contact sales link
- [ ] Help / docs link
- [ ] Money-back guarantee, restated

CTA means call to action: the button that tells the visitor what to do next.

Pro tip: look at the Linear, Notion and Figma pricing pages. They're well-known examples of a clear pricing page, so copy the structure (not the prices). The prices on those pages change; look at the current ones.

Step 4: Test your pricing: a 30-day plan

After launch, you collect data.

Weeks 1-2: Baseline

• Track conversion (visitors → trial → paid)
• Track the tier split (% Starter / Pro / Business)
• Do 1-on-1 calls with trial users and listen to their objections

Weeks 3-4: Iterate

• If conversion is very low, rethink the Starter price and the offer
• If almost everyone takes Starter, widen the gap (tighter Starter limits)
• If nobody goes for Business, improve the Business value statement
• The same sales objection comes up 3+ times? → adjust

Day 30: Review

Rough targets for a first comparison; set your own from your own data.

Step 5: Document your pricing decisions

Create a pricing.md file in your project:

# Pricing — Decision Log

## Current pricing (vYYYY-MM)
- Starter: $29/month
- Pro: $99/month
- Business: $299/month
- Annual: 20% off

## Reasoning
- Customer value avg $700/month (from 5 interviews)
- Pricing 15% of value = $105 target
- Rounded to $99 (charm pricing)
- 3x rule: $29 → $99 → $299 ✓

## A/B tests history
- [Date]: Tested $79 vs $99 Pro → $99 won on revenue
- [Date]: Tested $29 vs $39 Starter → $29 won on conversion

## Pricing changes
- [Date]: Initial launch
- [Date]: Added Business tier
- [Date]: Raised from $19/79/199 to $29/99/299

## Next review
- [Date]: Quarterly pricing committee

This becomes your pricing memory. A year from now, you won't remember why it's exactly $99.

Readiness checklist

• ✅ Customer value quantified ($X/month, with customer interviews as the source)
• ✅ 3 tiers designed with the rule of 3x ($29/$99/$299 or similar)
• ✅ Tier features mapped (what's in each, building up progressively)
• ✅ Annual discount calculated (15-20% off, clearly visible)
• ✅ Pricing page live with all 7 elements (hero, pricing cards, enterprise CTA, trust signals, FAQ, comparison table, footer)
• ✅ Money-back guarantee announced (30 days by default, only if you'll honor it)
• ✅ Price testing plan ready (sequential, not simultaneous)
• ✅ pricing.md written, with a decision log
• ✅ 30-day review scheduled (conversion, tier split, churn)
• ✅ Grandfathering policy decided (for future price increases)

If you're at 8 out of 10, you're ready to launch. At 6 or 7, finish what's missing. If you're under 6 out of 10, go back to Step 1 of the practice.

Tools and resources

• Stripe Pricing Page: an example of pay-as-you-go pricing: a fee on each payment (a percentage plus a fixed amount), with no monthly fee
• Linear Pricing Page: a clean tier example
• Notion Pricing Page: a freemium + tiers reference
• Prices for AI assistants and APIs: current numbers for your margin math
• Books: "Monetizing Innovation" (Madhavan Ramanujam and Georg Tacke), "Pricing Done Right" (Tim J. Smith)

Key takeaways

Pricing isn't just "what it costs." It's a signal of value (what the customer expects), a filter for customers (who you let in) and your source of revenue (how much you make). Three jobs in one number. Underpricing "so at least someone buys" makes a business hard to sustain.

Value-based pricing is the right path for AI. Cost-based pricing leaves money on the table; competitive pricing ignores what sets you apart. Quantify the value ($X a month in time saved, revenue uplift or cost reduction) and charge 10-20% of it. That's a common approach to AI pricing in 2026.

Always 3 tiers, with the rule of 3x ($29 → $99 → $299 as an example). A single tier loses part of the market at both the top and the bottom. Pro is the main offering, and most customers are likely to pick it. Starter is the entry point, Business is the premium option.

A hybrid beats a pure model. Pure usage-based pricing scares customers ("how much will this cost?"), and pure flat pricing doesn't scale. A base subscription + overage for heavy users = predictable and scalable. The cell phone plan pattern works for AI.

Annual + grandfathering = a strong retention tool. An annual discount brings cash flow, commitment and usually lower churn. Grandfathering existing customers when you raise prices brings loyalty and fewer waves of cancellations.

A/B test prices sequentially, not simultaneously. Never show different prices at the same time (PR risk). 30 days at price X, 30 days at price Y, then compare revenue per visitor (not conversion alone).

Related lessons

• Pricing: the previous lesson, on pricing a service by its value to the client
• Real monetization case studies: pricing examples
• Unit economics of an AI stack: LTV, CAC (customer acquisition cost) and payback period in detail; a library lesson, optional
• Funnel design for AI products: the pricing page inside your funnel; a library lesson, optional

Next lesson

→ How to find your first clients through people you know