How Should Startups Approach Conversion Rate Optimization & A/B Testing

Thoughts on how to improve conversion rate by betting on more comprehensive tests & tradeoffs that come with this approach.

Jun 3, 2025
For most companies, when you’re talking about conversion rate optimization or a/b testing; the first thing that comes to mind is changing button colors. For larger enterprises, even the tiniest change can output tremendous value. However, startups have more aggressive growth goals, and their approach to optimization should be completely different.

What Should We Test

Optimization and experimentation within startups is all about taking risky bets, and pivoting quickly when the bets fail. Here are some areas of your inbound initiatives you might want to think about optimizing:
  • Conversion Funnel
    • Should we introduce a quiz instead of a contact form?
    • Are there any other intermediary actions we can introduce to improve lead gen?
    • Which steps did our current customers take when they were still prospects? How can we optimize & scale that funnel?
  • Landing Pages
    • Not just button colors, but comprehensive tests where you’re not isolating a single variable. Experiment with a completely fresh design, messaging and layout. This can usually yield to quick wins; conversion rate lifts of >50% are not uncommon.
  • Ad Creative & Targeting
    • The whole point is to test at scale without sacrificing some of the statistical significance. Test 10 combinations of creatives and targeting tactics simultaneously if your budget allows. Then, take the top performing combinations and scale them!
 

Tips & Tricks

  1. Take big bets!
  1. Pivot quickly if the bets don’t pay off- this is often the most difficult part.
  1. Push back on experiments and optimization that won’t deliver lift rates of less than 15-20%.
  1. Document wins & losses to guide future experiments.
  1. Choose success metrics before running tests and ensure you have reliable means to track these metrics. It’s a common occurrence where halfway into an experiment, teams find out that they have no way of tracking their data, or their data sources are not reliable.
  1. Don’t wait for 95% confidence intervals, but also don’t make decisions on sample sizes that are too small to be meaningful.