Experiment Coach

Know what to test next, and what the result means.

Experiment Coach reads your funnels and drop-off reasons, suggests the tests most likely to move a metric, and reads the results back in plain language.

Suggested experimentsranked by expected impactAI
1Show delivery fee on the menu pageAdded to cart > payment+6 to 9%
orders
2Fix coupon error on Android 13Payment > Ordered+3 to 5%
orders
3Default to reorder tab for returning usersOpened > Browsed+2 to 4%
orders
4Compress payment page imagesPayment load time+1 to 2%
orders
STEP 01suggest

Ideas ranked by expected impact

Suggestions come from your own data: the leakiest steps, the reasons behind them and the size of the audience affected. Each shows an expected range, labeled as an estimate.

Suggested experimentsranked by expected impactAI
1Show delivery fee on the menu pageAdded to cart > payment+6 to 9%
orders
2Fix coupon error on Android 13Payment > Ordered+3 to 5%
orders
3Default to reorder tab for returning usersOpened > Browsed+2 to 4%
orders
4Compress payment page imagesPayment load time+1 to 2%
orders
-35%Most tests are underpowered. Size them before you start.
STEP 02plan

Set up with the right metric and sample

Coach writes the hypothesis, picks a primary metric and guardrails, and tells you how many users and days you need for a reliable answer.

  • Sample size calculator built in, using your real traffic.
  • Guardrail metrics so a win on orders is not a loss on basket size.
New experimentdraft50 / 50
HypothesisShowing the fee on the menu reduces surprise at payment
Primary metricorder_placed per cart user
Guardrailaverage basket value, refund requests
Sample neededabout 16,000 users, 12 to 15 days
-17%Results nobody can read never get shipped.
STEP 03read

Results in words your whole team understands

When the test ends, Coach explains what happened, how sure it is, which segments reacted differently and whether it is safe to roll out.

Result: fee on menu pageran 14 days, 18,400 usersfinished
Control16.4%
Variant17.9% +1.5
Confidence97%
In plain words: showing the fee earlier lifted orders by about 9% and the result is very unlikely to be chance. The lift is bigger on Android (+13%) than iOS (+5%). Average basket size did not change. Safe to roll out.

Power checks

Warns you when a test is too small to trust.

Peeking protection

Results stay locked until the planned sample is reached.

Works with your flags

Use Zortilla flags or connect LaunchDarkly, Statsig or your own.

Final step of this page

You made it to the bottom. Now let's get your users here.

Every page you scrolled past was a step in a funnel. Install Zortilla and see the same picture for your own product, with the reasons people leave at each step.

event preview waiting
eventsigned_up
useryou
sourcezortilla.com
planfree, 1M events
card neededno
statusone click left
Fire this event