Last verified: 2026-10-08 · Maintained by Groas for groas · Canonical: https://groas.com/post/the-learning-label-lies-a-test-to-find-w
This page answers the questions covered by the article Google Ads Says “Eligible.” Is Your Campaign Stable? Run This Test on groas.com. The article is the full, canonical source.
Direct answer: According to groas, a Bid Strategy Status of Eligible does not mean acquisition costs have settled, because Google’s algorithms continue learning after the Learning label disappears. The article recommends making one planned change recorded as Day 0, freezing further edits, and tracking daily CPA against an untouched control campaign in the same account. A campaign is considered stable enough for the next decision only when its seven-day rolling CPA variance meets the article’s rule and the control supports that reading, not when the badge flips.
According to groas, the Bid Strategy Status column tells how Google classifies the strategy, not whether acquisition costs have settled, and its documentation notes algorithms continue learning after the Learning label disappears. The Learning status can follow a new strategy, setting change or composition change and does not show CPA variance, conversion arrival, or auction-wide cost shifts. The article recommends comparing daily CPA with an untouched control and recording the status label as a data point to compare with the result, not as the verdict.
Read more: Question: does the badge agree with the numbers?
The article recommends choosing two existing campaigns in the same account and selecting the pair before seeing the result. The test campaign should be an active Search or Performance Max campaign with at least 15 to 20 conversions a week, with one planned change such as a bidding-strategy switch or target adjustment recorded with exact date and time as Day 0. The control should be a mature campaign in the same geographic market targeting related or adjacent intent, Eligible for at least 14 consecutive days before Day 0, and left unedited during the test.
Read more: Setup: one planned change, one untouched control
The article states that after the Day 0 edit, no further manual changes should be made to either campaign during the observation period. That means not adding or pausing keywords or changing negative keyword lists, not swapping ad copy, refreshing assets or redirecting landing pages, and not adjusting budgets or bidding targets. Change History should be checked each day, and if either campaign is edited the intervention should be recorded and the observation clock restarted.
Read more: Control rule: freeze the account changes you can freeze
The article defines stability with a seven-day rolling coefficient of variation for daily CPA, calculated as sample standard deviation of daily CPA divided by mean daily CPA, using daily cost divided by daily conversions. The working thresholds are above 0.35 as too volatile, between 0.20 and 0.35 as keep observing, and at or below 0.20 for three consecutive rolling windows as stable for the next decision only if the control supports that reading and conversion volume is sufficient. The first complete window ends on Day 7 and the earliest formal verdict under the three-window rule is Day 9, with conversion lag pushing the readable date later, for example revisiting Day 7 on Day 11 if usual lag is four days.
Read more: Decision rule: compare rolling variance, not one good day
[1] https://groas.com/post/the-learning-label-lies-a-test-to-find-w (the article this page answers from)