Go to app

Test ideas: what to test first

Good first experiments, from cover letter length to rates and boosting, and how to keep each test clean so you can trust the result.

Written by the EarlyBird teamBETA

Last verified Nov 4, 2026

Watch the video walkthrough (opens in a new tab)

The best first test is one change you'd otherwise argue about. Pick something that affects whether clients open your proposal or answer it, change only that, and let the numbers decide.

Ideas built into the app

Click View test ideas in Experiment settings to see these while you set up a test:

  • Bold hook vs Regular text vs Emojis headline: same letter, different first line.
  • Template vs AI cover letter: a fixed Template against Enhanced AI. See Template vs Enhanced AI.
  • Freelancer connects vs Agency connects: bid from your own connects or your agency's. See Choosing which profile and connects to bid with.
  • Short cover letter vs Long cover letter: a few lines against a full pitch.

More ideas

TestChange only this in each variantWinning metric to use
Your rateTerms settings: profile rate vs a constant rate vs a share of the budgetReply rate
BoostingBoosting on vs off, or Smart vs ConstantView rate
AI instructionsThe bracket instructions in your Enhanced AI templateReply rate
Screening answersAnswers: AI answers vs your default answersReply rate
Specialized profileBid using this profile: general vs a specialized profileReply rate

As a rule of thumb, things the client sees before opening your proposal, like the first line and boosting, move the View rate. Things they read inside it, like the letter, rate and answers, move the Reply rate.

Keep the test clean

  • Change one thing, two at most. If variants differ in several ways, you can't tell which change made the difference.
  • Copy the control. Use the copy icon on Variant A to make an exact copy, then change the one setting you're testing. Variant B starts from the default settings, so it differs in many ways at once.
  • Send enough proposals. Aim for at least 10 proposals per variant. More is better: bigger tests give more reliable results.
  • Check each variant before you start. Open each one and confirm only your test setting differs.

After a test

When a winner is applied, it becomes your bidder's settings. Start the next test from there, so each experiment builds on the last. See Reading results and picking a winner.