How to Price Generative AI Products

In a marginal cost world, willingness to pay is king

Midjourney prompt: "Lemonade stand, watercolor, --c 100" (remixed and Photoshopped for final results)

Sponsored By: The Commons

This essay is brought to you by The Commons. Want to invest in your team and retain your top talent? Check out The Commons - a private community of 1000+ tech operators across 400+ companies.

It seems like a new generative AI startup is calling me every other day with the same question: how should they set their prices? 

Helping startups set their prices is not exactly new to me—I’ve spent years thinking about pricing in my work helping software and internet companies monetize their products. However, the recent uptick in questions from generative AI startups is worrying. Generative AI introduces a problem into software pricing that startups previously didn’t have to deal with: marginal cost.

Marginal cost (of running large language models, for example) breaks the startup product-launch playbook of “engagement first, monetization later.” Freemium is now a dangerous acquisition model. Subscriptions might be highly unprofitable. I hear plenty of doomsday scenarios. Still, if AI startups follow best practices, then pricing their products may not be that scary after all.

Pricing costly software

Let's address the cost issue right out of the gate. How much should you consider your costs when setting your prices? Pretty much not at all. 

Imagine I run a lemonade stand. I'm trying to set the price for a glass of lemonade. Now, I'm not just a lemonade entrepreneur but also a mind reader. My superpower is that I know exactly what my customers are willing to pay for my lemonade. There are 10 people waiting in line for my lemonade: the first customer's willingness to pay—or the maximum price they’re willing to pay for a product or service—is $9, the second customer's is $8, and so on and so forth, until we get to the last customer's willingness to pay of $0 for my lemonade. I have perfect market knowledge. So what's the optimal price for my lemonade?

Assuming that I can't charge different customers different amounts, the optimal price is $5. You can do the math yourself, but if I charge $5, five customers make a purchase, and I pocket $25.

What if I told you that my lemonade costs me $1 per glass to make? Does that change your answer? It shouldn’t. The optimal price is still $5. What if the lemonade cost $4.50 per glass to make? You might tell me I have a bad business, but the optimal price is still $5. Cost has no bearing on what my price should be.

Unlock your team's full potential with The Commons - a premier learning community that gathers 1000+ insightful tech operators from 400+ trailblazing companies. 

Did you know 94% of your team would stick with you longer if you invested in their growth? Say goodbye to ineffective training programs and shift to their empowering Sprints, designed to boost your Product, Strategy & operations, Growth, and Strategic Finance teams.

Companies like ResQ, Float, and Thumbtack are already harnessing the power of collective learning and growth with The Commons. Now, it's your turn. Their cohorts are curated and highly limited, but you can get priority access here.

An astute reader might point out that it makes a difference whether we are trying to optimize revenue or profit. Let's take that objection seriously. I did the math, and your costs have to be $2 per glass before our "revenue optimal price" and our "profit optimal price" diverge. Put another way, you have to have 60% or lower gross margins before cost starts making a difference.

Does your generative AI product have 60% or lower gross margins? I hope not.

Why willingness to pay trumps competitive intelligence

Another common mistake I see AI startups make is looking over their shoulders to see what everyone else is doing. Instead of focusing on the unique value that your product provides, you spend all your time Googling competitor prices and scraping pricing pages. Competitive intelligence is almost always insufficient for answering your pricing question. 

Back to lemonade. Let's use the same example—10 customers with willingness to pay ranging from $0 to $9. If I tell you that other lemonade stands on the block are charging between $3 and $7 a glass, where should you price your lemonade? 

The optimal price is still $5 (remember, I'm a mind reader who already knows every customer’s willingness to pay). What if I didn't have my customers' willingness to pay? Competitive intelligence still doesn’t help. Should I price my lemonade in the middle of the market? Should I price above or below the market range? I know nothing. 

I hear you complaining already. "What if a new competitor comes in and starts taking all of my customers?" That, of course, can happen, but you still need to listen to your customers. If you're out there talking to them all the time, you should see your customers' willingness to pay drop as new competitors enter the market or drop their prices. Don't just lower your prices because some ankle-biting competitor slashed theirs.

The right way to price

Understanding willingness to pay is the first step to coming up with the right pricing strategy. Startups should conduct at least 10 customer interviews per month to gauge willingness to pay. Once you have a sense of which customers get value from your product, you can design a monetization strategy that encourages every customer to pay their fair share.

I like to start by picking a good price metric. The price metric is essentially what you are charging for: users, API calls, gigabytes of storage, compute. The right metric correlates well with willingness to pay. Price metrics do not have to correlate to costs, which are largely irrelevant when determining your price. This is a mistake many AI companies are making right now.

Next, you’re going to want to think about the price structure—how your price metric relates to time and volume. Do you pay in advance each month? Do you buy blocks of users? Are the first 1,000 API calls included? What about volume discounts? You’ll want to bend your price curve to match your customers’ willingness to pay at different volumes. Try not to have an overly complicated structure; after all, more complicated structures are harder to sell and implement.

The last step is setting a price, often called the price level. But before you get here, make sure you’ve done enough customer interviews. What about surveys or sales testing? Leave the heavy A/B testing software or giant market research studies alone for now. You can triangulate to a “good enough” answer using some of the methods I’ve described above—and getting the metric and structure right are way more important than polishing the price level, which can be adjusted on an annual basis.

How to get started with pricing

I’ve been talking about measuring customer willingness to pay as if it’s something you can do with a measuring tape. In practice, you need to use one of many “pricing methodologies.” One of my favorite methods is the “Van Westendorp,” a technique introduced by Dutch economist Peter van Westerndorp in 1976 that helps determine consumer price preferences. Ask customers what a reasonable or bargain price would be for your product, as well as an expensive and prohibitive price. By getting three or four data points per customer (some methods also ask for “too cheap”), you can approximate a demand curve for a whole segment of customers.

There are lots of other fancy ways to determine willingness to pay, but most startups can fall back on the good ol’ Van Westendorp. It helps to have one person own pricing decisions, because quick iteration is crucial.

AI should be treated with the same good pricing hygiene as any other product—marginal costs or not. Focus on willingness to pay, not costs or competition. Set your price metric, structure, and level by interviewing customers, and consider using simple price studies, like a Van Westendorp, to judge willingness to pay. 

And lastly, throw out that “monetize later” playbook. Place pricing first, where it belongs.


Ian Clark has over a decade of experience helping software and internet companies monetize their products. His clients include companies such as Y Combinator, LinkedIn, Eventbrite, and Cloudflare. He is the author of a book on monetization.


Like this?
Become a subscriber.

Subscribe →

Or, learn more.

Thanks to our Sponsor: The Commons

Thanks again to our sponsor The Commons, an exclusive community of 1000+ tech professionals across 400+ companies. To increase employee retention, it's super important to invest in their career growth – something which 94% of employees affirm. With often brutal and inadequate training programs, tech companies must invest in progressive learning and mentorship avenues like The Commons.

Their October cohort is nearing capacity, but you can get priority access as an Every subscriber.

Thanks for rating this post—join the conversation by commenting below.

Comments

You need to login before you can comment.
Don't have an account? Sign up!
Shawn Cho 7 months ago

Loved this post - thanks for writing it.

Could you provide more explanation regarding optimizing for revenue vs. profit, and the divergence of "revenue optimal price" and "profit optimal price"?

Every smart person you know is reading this newsletter

Get one actionable essay a day on AI, tech, and personal development

Subscribe

Already a subscriber? Login