Pricing experiments, with the results kept

Solo buildersIndie hacker testing prices on a small SaaS

Over four months we ran three pricing page tests: annual-first, a cheaper starter plan, and a free trial without a card. When we asked an agent to plan the fourth test, it proposed the card-less trial again, which had doubled signups and halved paid conversions.

How do I keep past pricing test results so my AI agent doesn't suggest the same test again?

After each test, save the result in words with ArcRoom's remember: what changed, how long, what moved. Record the call you made from it with arcroom decide. When a prompt mentions pricing or the pricing page, recall hands over past tests and their outcomes, so the agent plans a test you haven't run yet. ArcRoom doesn't connect to your analytics: you type the numbers in.

Also asked as: My agent proposed a pricing test that already hurt conversions. How do I stop that? · How do I give an AI agent the history of my pricing experiments? · Can an AI agent remember what my last pricing tests showed?

The work

After each test, save the result in words with remember: what changed, how long, what moved.

ArcRoom keeps
  • Decision
  • Saved decision
  • Recall
The next agent gets

Remembered (Aug 30): test 3, trial without card, 3 weeks.

Step by step

  1. After each test, save the result in words with remember: what changed, how long, what moved.
  2. Record the call we made from it with arcroom decide.
  3. When a prompt mentions pricing or the pricing page, recall hands over past tests and their outcomes.
  4. The agent plans a test we haven't run yet.

What the agent sees

Remembered (Aug 30): test 3, trial without card, 3 weeks.
Signups 2.1x, paid conversion 0.5x. Net: worse.
Decision: keep the card on trials.

Compared with today's tools

Notion or a spreadsheet

Log each experiment in a table and paste the relevant rows into the agent before planning.

Setup
30 min
Ongoing
paste the history into each planning session, ~5 min

ArcRoom

Every agent reads the team playbook before it works and writes to it when it finishes. Nothing to update by hand.

Setup
2 min (npx arcroom init)
Ongoing
None

A spreadsheet is the better record with real numbers; ArcRoom's edge is that the agent gets the result without anyone pasting it.

Where it stops

ArcRoom doesn't connect to your analytics. You type the numbers in, and they're only as good as what you saved.

Try it in your project

Works with the agents your team already uses, and any agent over MCP or the HTTP API.

More in solo builders