I launched AdLoft at $19/month and left it there for eight straight months while MRR crawled past $2,100.
The first three mistakes that kept revenue flat
- I anchored to competitor hobby prices. Every other AI tool sat at $19–$29, so I copied them without testing what my specific users would pay for ad creatives.
- I treated pricing as a set-and-forget setting. I updated the landing page copy six times but never touched the pricing table again after week one.
- I optimized for sign-ups instead of lifetime value. Cheap plans brought in 140 trials a month, yet only 11% converted and most canceled before month three.
The experiments that finally moved the needle
Month nine I ran three parallel pricing tests with existing users and new traffic. One cohort saw a $49 plan, another $79, and the last a usage-based tier. The $79 option won: average revenue per user jumped 3.4× and churn dropped to 4% monthly. During the same period I also shipped a background remover for ecommerce that let users generate clean product shots instantly; the feature justified the higher price for many store owners.
- Usage data beat gut feel every time. Once I pulled actual render counts from the dashboard, it became obvious customers who created 30+ ads per month would happily pay triple.
- I stopped selling seats and started selling outcomes. The new messaging framed AdLoft as “one profitable ad creative per day” rather than “AI image generator access.”
- Annual prepay unlocked the final 28% lift. Offering 20% off for yearly billing pushed 41% of customers onto annual plans within six weeks.
What I track now instead of vanity metrics
Every quarter I rerun the same three-cohort test on fresh leads. I watch two numbers: willingness-to-pay median from interviews and realized LTV from the last pricing cohort. When the gap between them shrinks below 15%, I know the price is close to right.
I’m Didar, the founder of AdLoft — an AI ad creative platform turning product photos into professional ads. I write here about the stuff I’m building and what’s working.