Microsoft Ads optimization experiments: what to test first?

Short answer: Microsoft Ads optimization experiments became generally available when Microsoft announced them on 30 September 2026. They let you A/B test bidding, targeting or creative changes in Search, Shopping, Audience and Performance Max campaigns. Test your biggest planned change first, usually the bidding strategy, in the campaign with the most conversions. Change one thing at a time.

What changed in Microsoft Ads on 30 September 2026?

Microsoft Advertising announced on 30 September 2026 that optimization experiments are generally available for Search, Shopping, Audience and Performance Max (PMax) campaigns.

An optimization experiment is a controlled A/B test. You change one setting, such as bidding, targeting or creative, and measure it against the setup you run today. Results for all your experiments appear on one unified page.

When a change wins, you have two options: apply it to the original campaign, or use it to create a new campaign.

The same update raised the LinkedIn company list limit from 1,000 to 10,000 company names per list. You use these lists with LinkedIn profile targeting in Search and Audience campaigns. European advertisers should know about one catch, which I cover further down.

Why this matters on a small budget

Without a test, the only way to try a new bid strategy is to switch the whole campaign and hope. In a controlled test the original setup runs over the same dates as the change, so season and competitors affect both sides equally. If the change fails, it costs you part of the traffic, not a month of sales.

What should you test first?

Start with the change you were going to make anyway, the one that moves the most money. In most accounts that is bidding. Creative tests often show a difference in CTR (click-through rate) quickly, but CTR is not revenue. Judging creative on sales takes longer.

Change to test	Test it when	Judge it on	Skip it when
Bid strategy, e.g. manual CPC (cost per click) to target CPA (cost per acquisition) or target ROAS (return on ad spend)	Tracking is reliable and the campaign gets conversions every week	CPA or ROAS, together with conversion volume	The campaign gets only a handful of conversions a month
Target level, e.g. a stricter target CPA or a higher target ROAS	Volume is fine but each sale costs too much	Conversion value minus spend, not ROAS alone	Your conversion values don’t reflect margin yet
Targeting: locations, audiences, LinkedIn profile targeting	You suspect part of the budget reaches the wrong regions or people	CPA plus lead or order quality	You need LinkedIn profile data to reach users in the EEA (European Economic Area), the UK or Switzerland
Creative: headlines, descriptions, images	Bidding is stable and you have a genuinely new message, such as price, delivery or an offer	Conversion rate first, CTR second	The new ad only rephrases the old one

If you plan a target ROAS test, check your conversion values first. A test built on revenue that ignores margin can easily pick the wrong winner. The setup logic is the same as in my guide to Google Ads conversion value setup for profit.

How to run an optimization experiment step by step

Menus change, but the process doesn’t. I follow the same steps for experiments in Google Ads.

1. Check tracking. Make sure the UET tag (Universal Event Tracking, Microsoft’s conversion tag) still records purchases or leads once visitors have seen your cookie consent banner. In EU accounts, consent is where tracking breaks most often without anyone noticing.
2. Pick one campaign. Choose the one with the most conversions, not the one that worries you most. Low-volume campaigns rarely give a clear answer.
3. Write the hypothesis and the decision rule. For example: target ROAS will cut spend without losing revenue; I apply it if ROAS stays above 400% and monthly revenue drops by less than €1,000. These numbers are illustrative. Set your own before you see any data.
4. Change one variable. If you switch bidding and rewrite ads in the same experiment, you won’t know which change did the work.
5. Leave both sides alone. Don’t edit budgets, add keywords or change ads while the test runs.
6. Read the result on the experiments page. Judge it on the metric you wrote down, not on whichever metric looks best.
7. Apply or discard. Apply the winner to the original campaign or use it to create a new one. A losing test costs you part of the traffic, not the whole campaign.

How long should the test run?

Long enough to collect a meaningful number of conversions on each side. There is no universal number, so calculate it: weeks needed = conversions you want per side divided by conversions each side gets per week.

Illustrative example: say the campaign gets 40 conversions a week and each side of the test gets roughly half the traffic. That gives you about 20 conversions per side per week. If you want at least 60 per side before you judge, plan for at least three weeks.

Run full weeks so the test covers both weekdays and weekends. If your customers usually buy several days after the click, wait that much longer before you read the result.

What I see in the accounts I run

Across the 500+ campaigns I’ve launched in more than six years, bad changes are rarely what costs the most money. The expensive ones are changes nobody can evaluate afterwards. A new bid strategy goes live in the same week as a price change and a new landing page. Conversions move, and three months later nobody knows what helped.

With Microsoft Ads, this problem is usually worse. In the European accounts I audit, it is usually the second channel. Someone copied it from Google Ads at some point, then left it alone because the volume seemed too small to matter. Settings chosen for Google stay there, untested, for years.

That is where experiments earn their place. Testing one change against the original is the cheapest way to learn whether settings chosen for Google actually suit Microsoft’s audience.

One habit I keep in every account: before a test starts, I write the decision rule into a shared note with the start date, the metric and the threshold. That way nobody, including me, can read a win into noise after the fact.

Does the bigger LinkedIn company list help EU advertisers?

For B2B (business to business) advertisers, the jump from 1,000 to 10,000 company names per list looks like the bigger news. You can now upload a full account-based target list for LinkedIn profile targeting in Search and Audience campaigns.

The catch: Microsoft states that the LinkedIn-based consumer data excludes users in the EEA, the UK and Switzerland. If you sell to companies in Slovakia, Czechia, Austria or anywhere else in the EEA, that data doesn’t cover your buyers there. The bigger list only helps in markets outside those regions, for example if you also target the US.

If your B2B market is European, LinkedIn’s own ad platform is still the direct way to target by company and job title. I cover which formats still make sense on a limited budget in LinkedIn Ads on a small budget.

When is an experiment a waste of time?

• Too few conversions. With a handful of conversions a month, a test can run for months and still not tell a real result from chance. Make the change directly and watch it closely, or test in a bigger campaign.
• Conversion values that ignore margin. A target ROAS test on raw revenue can pick as the winner the change that sells more of your least profitable products.
• The problem isn’t the ads. No bidding test will fix a slow product page, a price above competitors or a checkout that breaks on mobile. Fix the site first.
• Stopping early. If you end a test the day one side looks better, you can easily apply a change whose lead was just noise.
• Testing only in an unusual period. A test that runs entirely during a sale or a stock-out tells you about that period, not about normal weeks.

What to do this week

Open your Microsoft Ads account and pick the campaign with the most conversions. Name the one change you keep postponing, which is most often the bid strategy. Check that UET records conversions, write down the decision rule, and run the change as an experiment instead of switching the whole campaign.

If you’d rather have someone set up and read the tests for you, that is part of my work as a PPC (pay-per-click) specialist always inside your own accounts. You can start with a free advertising audit or look through all my PPC and performance marketing services. Email me and I’ll reply within 24 hours.