New format
A/B testing ads in Yandex Direct and Google Ads
A split test in paid search puts two (or a few) creatives or settings on comparable traffic so you can pick a winner on money metrics — not who won the CTR race.
Both Direct and Google Ads can rotate ads and run campaign experiments. Labels in the UI shift over time; the loop doesn’t: equal chance → enough data → decide on CPA or ROI. Landing-page tests live in the conversion optimization piece.
Why bother testing ads
Same query, different copy — and you can get different CTR, bounce, and cost per lead. Skip the test and you’re steering by gut or by one lucky week.
Platform rotation helps share impressions fairly. Reports and experiments go further: strategies, negatives, even bid logic.
Common test surfaces:
- headlines and descriptions
- sitelinks and callouts
- offer and USP
- landing pages (one change at a time)
- bids and strategies through an experiment
What Direct already does for you
Several ads in one group: after stats build up, the system often leans impressions toward the more clickable ones — gradually. Exact toggle names live in the current account UI.
Early on, variants usually share traffic. Don’t kill the “loser” after ten clicks.
Working habits:
- two or three variants, not a dozen
- one variable per test
- wait for real volume
- judge by goals, not CTR alone
Google Ads: rotation and experiments
Google long offered rotation modes — push the winners, or share more evenly. Names change; the choice is still “give everyone a shot” vs “favor what already works.”
Campaign experiments (the old drafts-and-experiments path) split traffic so you can compare bids, keywords, negatives, and strategies. Use them when a second ad isn’t enough.
Reading the reports
Direct: Report Wizard — cut by group and ad ID or text; pull clicks, conversions, CPA. Google Ads: ad reports and experiment reports.
Money beats vanity: CPA, ROAS/ROI, qualified leads, closed deals. Mind attribution — last click can undervalue upper-funnel ads.
Minimum columns:
- impressions, clicks, CTR
- conversions and CPA
- spend
- bounce / depth in analytics when you can
- search and display viewed apart
Not enough data yet
If a single ad is too thin, roll up carefully: compare copy only where the one variable actually differs. Don’t mash unrelated group themes into one “best text forever” ranking.
Broken UTMs and missing ad IDs wreck summary tables. Templates are covered in the UTM article.
Stop the test early if:
- conversions are still in single digits
- the week had weird demand
- you also changed bids and keywords
- search and display are mixed in one view
How to compare without fooling yourself
Same headline and keywords, only the body changes — grouping by text is fair. Headline plus different neighbor-group themes — “best text overall” will lie to you.
Keep search and display separate: CTR and traffic quality aren’t the same game. End each test with an action: pause the weak, scale the strong, or start a new hypothesis.
What to remember
A/B in Direct and Google Ads = rotation or experiments + money reports. One hypothesis, enough data, decide on CPA or ROI.
Low volume — roll up carefully or keep collecting. Don’t crown a one-day CTR champ.
FAQ
Can I just run two ads in one group?
Yes for headlines and body. For bids, strategies, or structure changes, use a campaign experiment if your account has it.
Does the ad with the best CTR always win?
No. A click magnet that burns budget on junk traffic loses to a quieter ad with a sane CPA. Score the business outcome.
How long should a test run?
Until conversions (or enough clicks) support a call — not a fixed “seven days from the handbook.” If traffic is tiny, don’t bother testing yet.
Can I change several things at once?
One hypothesis at a time: headline, or offer, or landing. Change everything and you’ll never know what moved.
Do I need UTMs?
For Metrica/GA4 side-by-side — yes, plus ad IDs in URL templates when the platform fills them.
Two ads fight on CTR — and you still crown a winner without CPA?
We’ll set one-hypothesis rotation/experiments in Direct and Google Ads and read reports by money, not a one-day CTR.
Discuss the task
