Reactive Pricing After Demand Signal
Let users pull pricing out of you, then validate willingness-to-pay and margins
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 88%
Gamma didn't pre-plan monetization; they focused entirely on activation, capped free AI usage with credits, and only built pricing once users flooded support demanding to pay. They then ran a quick Van Westendorp willingness-to-pay survey and conjoint analysis, anchored to the emerging ChatGPT $20/month reference price to minimize friction, and monitored for months to confirm the price was actually economical.
Origin
Grant Lee's account of Gamma's April-May 2023 pricing sprint; he notes the Van Westendorp survey is the same method ChatGPT used, as shared by the head of ChatGPT on the same podcast.
Core principles
- 01Willingness to pay is a second PMF checkpoint alongside organic growth
- 02A capped free tier (credits) surfaces genuine demand when users ask to pay
- 03Anchor to a familiar reference price (ChatGPT's $20/month) to reduce decision friction
- 04Keep V1 pricing simple (one plan) rather than overthinking it
- 05Pricing must be economical: verify you actually make money and can reinvest the margin
How to run it
- 1
Cap the free experience to surface demand
Give new users a fixed allotment (Gamma gave 400 AI credits) that cuts off, so real demand shows up as users asking how to buy more.
Pro tip Support-channel volume (Intercom blowing up) is a strong organic willingness-to-pay signal.
- 2
Run a Van Westendorp willingness-to-pay survey
Survey early users to understand overall willingness to pay across price points.
- 3
Layer in conjoint analysis
Use conjoint analysis to understand which features and elements users actually value, informing packaging.
- 4
Anchor to a familiar price and ship a simple V1
Land on a price near the familiar reference (~$20/month) with a single plan to minimize friction, then ship it.
Pro tip Default to removing friction and making pricing easy to understand over clever multi-tier schemes at V1.
- 5
Monitor margins for months
Track whether the price is economical: at this price point are you actually making money with strong enough margins to reinvest in headcount and inference? Tweak later if not.
Watch out With AI inference costs, a price that converts well can still be unprofitable; verify real margins, don't assume.
In the wild
After the March 2023 AI launch shipped pre-revenue with a 400-credit cap, Intercom blew up with users asking to buy more credits. Over April, Gamma ran Van Westendorp and conjoint analysis on early users, anchored to ChatGPT's ~$20/month, and shipped one ~$20 plan by end of May 2023.
→ Within a couple of months Gamma hit $1M ARR and became profitable, confirming the price was economical and could fund reinvestment.
Common mistakes
Assuming an AI product is profitable without checking
Inference costs mean a well-converting price can still lose money; Gamma explicitly monitored for months to confirm real margins rather than assuming.
Overthinking V1 pricing complexity
Elaborate multi-tier schemes add friction; anchoring to a familiar reference price with one simple plan let users decide quickly.
Is it for you?
Best for
AI/SaaS founders who focused first on activation and now need to introduce pricing under real inference costs
Not ideal for
Businesses with long runway that can defer monetization, or enterprise sales requiring bespoke negotiated pricing
From the transcript
“we did run a form of van westonorp which is just understanding what is the overall willingness to play”
“We did kind of integrate some forms of like conjoint analysis”
“you're almost uh end up becoming anchored buy what does chatbt charge”
“at the price point are we actually making money”
From the episode
“Dumbest idea I’ve heard” to $100M ARR: Inside the rise of Gamma
Grant Lee (CEO)