Analytics & CRO · 03

One change at a time.

Redesigns change everything at once, so nobody knows what worked. We test one change at a time on [10]% of traffic first, call it when it has enough orders, and roll out only the winners.

  • Product pages
  • Offers
  • Cart
  • Checkout
Change per test
1
Of traffic to start
[10]%
Orders per variant to call it
[200]
Two identical pump bottles side by side, one maroon and one yellowTest E[14] · A vs B

Where it sits

  • Growth sprintIn it
  • Growth partnerIn it
  • LaunchIn it
In short

How does A/B testing work on a store?

An A/B test shows two versions of a page to different shoppers at the same time and counts which one wins more orders. We change one thing per test, start on [10]% of traffic, run it until each version has about [200] orders, and roll out only the winners.

Tests
Product pages, cart, checkout and offers
Per test
One change
Starts on
[10]% of traffic
Called at
About [200] orders per variant
Kept
Every result, in the test log
Who it’s for

A fit, or not yet.

Honest both ways: if it isn’t the right first step, we’ll say which one is.

A good fit if

  • You redesign, and can’t say what worked.
  • Your store has at least [1,000] orders a month.
  • You want every change to come with a number.

Not the right fit if

  • You have fewer than [1,000] orders a month: we test bigger changes, or fix tracking and speed first.
  • You want results in a week. Tests need [2] to [3] weeks.
Not sure? Start with the leak
What’s included

In every testing month.

Compare the plans
  • A test plan

    Hypothesis, change and deciding number, before it runs.
  • Design and build

    Each variant designed and built by us.
  • Traffic split

    [10]% first, then everyone once it’s safe.
  • Clean reads

    Enough orders per variant before we call it.
  • Rollouts

    Winners rolled out to all traffic.
  • The test log

    Every test, its number and what we learned.
How we run it

How a test runs.

Every test follows the same steps, so every result can be trusted.

  1. Week 1

    Plan

    The hypothesis, the change and the number that decides it.

  2. Week 2

    Build

    The variant built and checked on phones.

  3. Every Monday

    Read

    Orders per variant, and whether it’s ready to call.

    Monday report[Name] · Analytics
  4. Weeks 3–4

    Roll out

    Winners to everyone; losers logged.

Price and plans

Where it sits.

Compare every plan
  • Growth sprint

    Included
    [Price]per [6]-week sprint

    [3] tests, as the sprint’s one service.

    Start a sprint
  • Growth partner

    Included
    [Price]per month · [3]-month start

    [2–3] tests a week.

    Book a call
  • Launch

    Included
    [Price]per project

    [5] tests planned for launch.

    Plan a launch

Prices are per plan, not per service. Every plan includes the audit, the 90-day plan and the Monday report.

FAQ

Questions about A/B testing.

Not on the list? Ask on a call, or by email.

Enough for a test to reach about [200] orders per variant in [2] to [3] weeks. Below that we test bigger changes, or fix tracking and speed first.

Still have questions?Saimul Islam answers every one.

Your turn

Let’s grow.

Book a strategy call

[30] minutes, no slides. You leave with the first [3] tests we’d run.

Or write to [hello@domain.com]