Know when your test has really won.

Run the test with the tool you already use. Kissmetrics reads the variant, the lift and the certainty together, and tells you when it is too early to pick a winner.

Try it on your own data. Free to start, no card.

https://

Included in Growth, $99 a month. Funnels and Activity stay free.

Result
One page leads the baseline
95% certainty and positive improvement required
Participants
3,646
1,705 needed per variation
Conversions
1,840
first conversion after assignment
Checkout: one page vs three stepsCounts people
Sample online store · Sep 2 to Sep 29goal: Placed order
Variation
Participants
Conversions
Conversion rate
Improvement
95% improvement interval
Certainty
Three stepsBaseline
1,834
879
47.9%
One pageLeader
1,812
961
53.0%
+10.7%
+3.9% to +17.4%
99.9%
One page wins, and the sample is there. 53.0% of people placed an order against 47.9%: +10.7% at 99.9% certainty over 28 days. Each variation has more than the 1,705 participants needed to detect a 10% improvement.

Illustrative numbers, fully clickable. Nothing here queries your data.

Asked in plain EnglishExamples
ForQuestion and answer
E-commerce
Did the one-page checkout beat the three-step one?
Yes, and it is safe to ship. +10.7% at 99.9% certainty over 28 days, with both variations past the 1,705 participants the test needed.
SaaS
Is the pricing page test ready to call yet?
Not yet. Annual first leads by +12.4% but certainty is 81.7%, with 447 of the 6,544 participants each variation needs.
Media
Which paywall wording produced more subscriptions?
The value-led wording: +18.5% at 99.7% certainty, with both variations past the 10,060 participants needed.

Ready the day you install. It fills in as your data arrives.

What it is good for

The usual mistake in testing is calling a winner early. This report puts the lift next to the certainty, and when the sample is too small it says so. Kissmetrics does not split your traffic: you pick the property or event that holds the variant, and the conversion that counts. Because every event belongs to a person, someone who comes back on another device is still counted once, in the variant they saw.

  • Knowing when a result is safe to ship
  • Measuring lift on the order or the signup
  • Reading tests from any testing tool on the same people
  • Holding the line when a test is flat

Same report, different questions

SaaS

Judge pricing and onboarding tests on activation. It is slower, and it is the number that counts.

What you set up

Annual-first pricing page against monthly-first, goal Connected data source.

What you get to decide

Annual first leads by +12.4% at 81.7% certainty, with 447 of the 6,544 participants each variation needs. You hold the change until activation says it worked.

Result
Not enough time or data to call it
95% certainty and positive improvement required
Participants
899
6,544 needed per variation
Conversions
190
first conversion after assignment
Pricing page: annual first vs monthly firstCounts people
Sample B2B SaaS · Sep 8 to Sep 29goal: Connected data source
Variation
Participants
Conversions
Conversion rate
Improvement
95% improvement interval
Certainty
Monthly firstBaseline
452
90
19.9%
Annual first
447
100
22.4%
+12.4%
-14.4% to +39.2%
81.7%

Shape it to your question.

Start from the ready-made report, then change it to fit what you want to know.

Ready-made, or asked

Open the report as built, or ask for it in plain English and refine it with a follow-up.

Your own events

Every event and property comes from your own schema, never a fixed list.

Any group of people

Filter it by a population, such as paying customers or one plan.

Any date range

Pick the window once and every report on the page follows it.

In your inbox, on your schedule.

Save the report and send it daily, weekly on the day you pick, or monthly, to up to 10 people.

  • Email. Any saved report, on a schedule. Free includes 2 scheduled reports.
  • Slack or webhook. The same schedule, sent to a channel or to your own endpoint.
  • Export. The report and its data as a CSV, on Growth, with no history limit.
  • API. Pull the same numbers into your own tools, on every plan.
  • Pin. Keep it on a dashboard next to the numbers you watch.
A/B testsUpdatePinEmail
  • Weekly, Mondays, Last 7 days, to maya@acmeapp.com, sam@acmeapp.comRemove
WeeklyLast 7 daysEmails (blank: just me)Schedule

Read next

Metrics

Your KPIs, split by the channel that drove themSee Metrics

Funnels

Where people drop off before they paySee Funnels

Cohorts

Who keeps coming back, week after weekSee Cohorts

Try it on your own data. Free to start, no card.

Install with AI in about 5 minutes.

https://

No card required. Free plan. No developer needed for your website.