Launch offer: 7 days free, then your first month free on monthly plans with code SELLER30. Until 31 December 2026.Start free trial
SellerPilot
العربيةSign inStart free trial
AI agentsProblemsFeaturesPricingSolutionsResourcesSecurityالعربيةSign inStart 7-day free trial

Test an ads change before you roll it out

A/B tests make one change on a test group of Sponsored Products keywords or campaigns (or, for a budget test, SP, SB or SD campaigns), and leave a control group of the same size, paired by spend and split at random, as it is. The figures are worked out every day on settled days; the result is read on day 14, then every 7 days up to day 28. A winner keeps its change and can be rolled out to the control group; a loser goes back by itself.

Start 7-day free trialAll features
No card needed Reminder 2 days before the trial ends Pay only if you choose a plan

Test a bid, budget or top-of-search change against a matched control group before you roll it out.

  • Bids up or down and top of search on Sponsored Products; daily budget on SP, SB or SD
  • A control group paired by spend and split at random
  • Judged on settled days: extra sales at your target ACOS minus extra spend
  • Losers and no-difference tests are rolled back by themselves

How it works

  1. Plan the test

    Choose the change (bid up or down by a percentage, top of search plus points, or budget plus a percentage), which keywords or campaigns (profitable, costly or all), and how long (14 to 28 days by default). The draft shows every change, the balance of the two groups and the smallest effect it can detect. Nothing changes on Amazon yet.

  2. Start it

    With your confirmation the change is made on the test group. While the test runs, the autopilot leaves both groups alone.

  3. It is checked every day

    Difference in differences per pair against the 28 days before, on settled days only. The decision figure is your target ACOS times the extra sales, minus the extra spend. The result is read on day 14, then every 7 days up to day 28; a test that is clearly losing money stops from day 7.

  4. Winner, loser or no difference

    A test that is 95% or more likely better is a winner: the test group keeps the change, and you can roll it out to the control group. A loser, a test with no clear difference by its last day, or one losing money with 97% certainty from day 7 goes back to the old values by itself.

Why a control group

A bid raise in a busy week looks good and in a slow week looks bad, whatever it did. A control group in the same weeks takes the season out of the result, so what is left is the change's own effect.

What it does not do

Stated plainly, so you know before you start.

  • It changes nothing on Amazon until you start the test
  • It does not keep a losing change: the old values go back
  • Values someone changed by hand during the test are left as they are
  • Small tests cannot detect small effects: the draft says what it can detect

Questions sellers ask

How long does a test run?
At least 14 days by default and at most 28, then a decision. You can set 7 to 56 days.
Does the autopilot interfere?
No. While a test runs, the autopilot leaves both groups alone.
What happens to the control group after a win?
Nothing until you roll it out: rollout shows the exact new values first, then applies them after your confirmation.

Related features

  • PPC autopilot

    An autopilot for bids, budgets, negatives and harvesting that never goes past your limits.

  • Bid optimization

    Bid and budget changes every morning, each with the clicks, spend, orders and ACOS behind it.

  • Strategies

    Six ads playbooks: Launch, Rank, Growth, Profit, Defend and Liquidate, with an advisor per campaign.

Try SellerPilot on your own store

Connect Amazon in about five minutes. Your first data arrives within hours and your first suggestions the next morning.

Start 7-day free trialSee pricing

No card needed. Nothing to cancel.