The premium product had five plans, including 1 month, 3 months, 6 months, 12 months, and lifetime. PDF archives were priced by intuition rather than by a pricing model.

The team wanted to test a low-ticket entry point to reduce the barrier to first purchase.

Problem

Conversion was limited, and the pricing page was confusing. But a cheap plan could also cannibalize the main subscription. The question was: can a micro-plan grow revenue, or does it just move existing demand to a cheaper product?

Evidence

  • 6- and 12-month plans produced almost no sales: one sale in August.
  • Most sales came from the 1-month and 3-month plans; lifetime was occasional.
  • The $1/day experiment:
    • 10 purchases;
    • PayPal took roughly 30% on each micro-payment, leaving about $0.60 net per sale;
    • monthly subscription sales fell to 1 during the experiment, versus the usual 3–4;
    • net revenue was lower than before the experiment.
  • AI search was built, but it fired rarely and users barely saw it. Its cost was under 2% of revenue.

Options considered

  1. Keep all five plans. Avoid change; accept complexity.
  2. Create a separate Premium+ plan for AI. A new SKU with a higher price.
  3. Put AI only in 3+ month plans. Use it as an upsell to longer commitments.
  4. Three plans, AI included, with limits. Simplify the choice and let AI support conversion.
  5. Keep the $1/day plan. Use it as a low-ticket entry.

Decision

  • Killed the $1/day plan. The unit economics did not work, and it cannibalized the monthly subscription.
  • Reduced five plans to three: 1 month ($10.99), 3 months ($25.99), lifetime ($99.99).
  • Removed 6- and 12-month plans. They did not sell and added confusion.
  • Included AI in all paid plans instead of creating a separate Premium+ SKU. AI was an amplifier, not a product: it fired rarely and cost little.
  • Did not raise the monthly price yet. The team needed conversion data after the plan change before adding another variable.

Execution

  • Rolled back the $1/day experiment.
  • Rebuilt the pricing page with three plans, a clearer value description, and price-per-month framing.
  • Added AI limits to the 1-month plan (20 answers/day) and kept it unlimited on 3-month and lifetime.
  • Planned to measure conversion, plan mix, and AI usage in October before changing the price again.

Result

  • The $1/day hypothesis was rejected: fees and cannibalization made it worse than the standard subscription.
  • The 5→3 simplification shipped. October data was not final at the time of writing.
  • AI was launched as a feature inside paid plans, not as a separate product. Usage and cost were being measured.

This case is included because the decision quality matters more than a clean win. The failed experiment produced a clear rule: do not launch a micro-plan before calculating payment fees and cannibalization.

What I learned

  • Micro-payments can die on payment fees before they ever reach product-market fit.
  • A cheap plan can cannibalize the core subscription even when it looks like “growth.”
  • Fewer plans reduce decision fatigue; pricing is part of the product experience.
  • Do not create a new SKU before you know the feature is visible and valuable.
  • What I would do differently: calculate unit economics before the experiment, define a kill criterion, and test with a small cohort instead of launching to everyone.