Microsoft swapped its bidder days before Black Friday

On Wednesday, Microsoft Advertising announced that it has rebuilt the model behind every automated bid on its network. Not a new strategy or toggle, but the thing underneath all of them: the prediction of whether a click will turn into a sale and what that sale is worth.

Microsoft calls it “the biggest upgrade to bidding models in a decade.” It is rolling out automatically, account by account, starting now. Black Friday is November 27.

That timing is the whole review. If you run Maximize Conversions or Maximize Conversion Value on Bing and its partner network, the brain spending your peak week budget may change during that week, and the announcement gives you no way to know when. The upgrade might be good. But only one of the three headline figures describes the upgrade itself, and the practical advice buried in the post matters more than the launch line.

This is a docs-based review. I have not had an account switched over, and nobody outside Microsoft can see which accounts have it yet. What follows comes from Microsoft’s announcement by product manager Carlo Buchmann (November 18), Microsoft’s help page on the bid strategy report, Google’s help page on its August target change for contrast, and coverage from Search Engine Land, Search Engine Journal and MediaPost (all November 18) and PPC Land (November 19).

What it is

Every automated bid is a guess about the future. Microsoft’s post puts it plainly: “Every bid depends on a prediction of whether a click will become a sale, and how much that sale is worth. We rebuilt that prediction model with new user signals and a better way of valuing each conversion.”

That is the entire technical disclosure. The new signals are not named and the “better way of valuing” is not explained, so it is worth being precise about what Microsoft says the model does.

Three things. Bids hit your CPA and ROAS targets “more consistently” without giving up volume or value. Bids adjust “nearly twice as quickly” when you change a target. And budgets “pace more evenly throughout the day and campaigns ramp to full budget faster.”

Notice that two of those three are about time, not results. Faster reaction to a target change means less time spent bidding to your old instruction. Faster ramp to full budget means your money goes out the door sooner in the day. Neither comes with a baseline. “Nearly twice as quickly” is not expressed in hours or days.

What changed, and what the numbers measure

The post cites three averages, each footnoted to Microsoft internal data. They are not three measurements of the same thing.

The first is the real one. Advertisers on Maximize Conversion Value or Target ROAS saw “on average 10% more conversion value per click at a similar ROAS.” The footnote says global Search and Audience Network data, August 2026. That is the only figure the post presents as the effect of the new model.

The second says Search advertisers who switched from Target CPA to Target ROAS saw 8% more conversion value at similar ROAS, using August and September data. The third says advertisers who switched from Enhanced CPC to Maximize Conversions with a CPA target saw 18% more conversions at a similar cost per action, using June and July data.

Those two are strategy switch numbers. The 18% figure comes from a period before the August window behind the 10% figure, and the post does not say whether the new model was running for any of it. None of the three comes with a count of advertisers, a count of campaigns, or a definition of “similar.” MediaPost noted the same gaps. And averages across advertisers who chose to switch describe advertisers who chose to switch.

So read the post as two things stapled together. A model upgrade with one decent sounding number, and a sales pitch for value based bidding with two numbers you cannot check.

Linocut print of a farmer opening a sluice gate at dawn so water rushes out of a reservoir while distant fields at the end of the channel stay dry.

What works

Strip out the launch language and the post carries unusually direct bidding guidance.

The five conversion floor is the big one. Microsoft now says that if an account has at least five conversions, new campaigns should start on a conversion based strategy from day one rather than Enhanced CPC, because “your conversion history at the account level provides bidding signals to all the campaigns within your account.” Only brand new accounts under five conversions should start on Enhanced CPC. A media buyer quoted by PPC Land pointed out that five is “way below the old 15 to 30 rule of thumb.” Microsoft’s Ads Liaison Navah Hopkins, replying on LinkedIn as quoted by Search Engine Journal, said the system “is much more responsive now!”

The target change guidance is useful and specific. Large target CPA increases or ROAS decreases “open up new inventory,” so expect short term volatility. If you want to avoid it, move in 15% steps and review after one conversion cycle, meaning the time it usually takes your business to turn a click into a conversion. The changes Microsoft says cause the most volatility are creating a new campaign, switching bid strategy, and adding or removing conversion goals.

The evaluation rule is the one most teams break. Judge results after “ideally 30 conversions in 30 days,” wait one to two conversion cycles after a major change, and judge at the campaign or portfolio level, not the ad group or keyword. That is a polite way of telling you to stop reacting to Tuesday.

And there is a tool for it. Microsoft’s bid strategy report shows target versus actual CPA or ROAS, a conversion delay metric that tells you when recent data is fully attributed, and status flags including “Limited by budget,” “Limited by target,” “Limited by max CPC” and “Limited: low data volume.” One detail on that help page is easy to miss: the average target shown is “the weighted average of the targets the auto-bidding strategy used to manage spend,” not the average of what you typed in. That number is your best window into what the bidder is actually doing.

What breaks

Four things, in order of how much they will cost you next week.

You cannot see the switch. The post says improvements will “continue to roll out automatically to more accounts over time.” It does not say which markets or accounts get the model first, how to tell whether yours has it, or whether you can opt out. Search Engine Journal put the consequence plainly: campaigns “can start spending differently before you make any changes yourself.” In Cyber Week that means any change in pacing, CPA or ROAS has two possible causes, the holiday auction and a new model, and you cannot separate them.

Faster ramp is not free. “Campaigns ramp to full budget faster” sounds like good news until you think about a day with a hard cap. If your budget runs out earlier, you lose the evening. For retailers whose Black Friday orders peak late, earlier ramp could spend the day’s money before the best hours arrive. Microsoft says pacing is more even through the day, which should prevent that. Verify it. Watch hourly spend, not just daily totals.

The same target now means different things on two platforms. Here Microsoft takes a shot at a competitor without naming it. “Unlike other platforms, Microsoft Advertising’s bidding continues to allow campaigns to overachieve on target CPA or target ROAS efficiency, regardless of budget limited status.” Google went the other way on August 17. Its help page says budget limited campaigns on target strategies “now perform more consistently toward your bid target,” with the example of a $10 Target CPA campaign that had been delivering at $5 moving toward $10. So if you mirror targets from Google into Microsoft, by import or by habit, the same number now produces different behaviour. On Google, a loose target in a budget capped campaign gets spent up to. On Microsoft, Microsoft says the bidder may keep beating it.

The max CPC argument will tempt people into a bad Friday. Microsoft says keeping a Maximum CPC is “not recommended” because it overrides your CPA or ROAS target and causes advertisers to miss outcomes “even when the cap is above average CPC.” That may be true over a month. But a cap is the only hard brake many teams have on a click during an auction frenzy. Removing it on November 25 because a blog post said so is how you find out what an uncapped holiday click costs. The decision is fine to make. Make it on purpose, with a target you trust, not as a reflex.

Linocut print of two identical weather vanes on neighbouring rooftops pointing in opposite directions in the same wind while a person below looks up at both.

Who it is for

Value based advertisers with real transaction values are the clear winners on paper. Microsoft says “the greatest improvements come when advertisers optimize using actual transaction values, such as purchase revenue.” If your purchase conversion sends the actual order total, you are the audience the 10% number describes.

Lead gen teams are next, with homework. A form fill has no revenue at the moment it happens. Hopkins, as quoted by PPC Land, called sending an estimated value at submission and updating it later through the Conversions API or offline import “the ideal use case,” with the warning to keep the ROAS target “aligned with the conversion data you’re passing through.” If sales has not agreed what a qualified lead is worth, a ROAS target is a guess wearing a decimal point.

Small accounts get the most interesting change and the least protection. The five conversion floor lets them start on automated bidding sooner, but Microsoft’s own evaluation bar is 30 conversions in 30 days. An account that qualifies to start may need well over a month before its results can be judged by Microsoft’s own standard.

Pricing and limits

No new fee. The model sits inside existing automated strategies and you pay for clicks as before. What it costs you is attention.

The limits are the gaps. No named signals. No baseline for “nearly twice as quickly.” No sample sizes. No rollout map, no in account flag, no stated opt out. The post calls this “the first stage” and promises more automatic improvements in 2027, without saying what they are.

What to do this week

Draw a line on November 18. Annotate the date in every report and dashboard that touches Microsoft Ads. When someone asks in December why ROAS moved, the model change should be on the chart.

Freeze discretionary changes through Cyber Monday. Microsoft’s own list of volatility triggers is new campaigns, strategy switches and conversion goal changes. Do not stack any of them on top of a model rollout and a holiday auction. If you must move a target, move it 15% and wait a conversion cycle.

Open the bid strategy report for your top campaigns today. Note the weighted average target, the actual CPA or ROAS, the status and the conversion delay. That is your before picture. Check hourly spend on Black Friday against last year’s curve to see whether earlier ramp is eating your evening.

Audit conversion values before you touch ROAS. Pull five recent orders and compare the value Microsoft recorded with the real order total. Search Engine Journal suggests exactly this check. Fixed values make value based bidding a fancy volume strategy.

Decide on max CPC caps on purpose. List every campaign that still has one. Keep the caps through Cyber Week where a runaway click would hurt, and schedule a controlled removal test for December where Microsoft’s argument can be measured.

Stop copying targets from Google. For any campaign set mirrored across both platforms, note which ones are budget limited. Those will behave differently on each side under the same number, and your blended reporting should say so.

Put the strategy switches on the December list, not this week’s. If you are on Enhanced CPC with more than five conversions, Microsoft’s case for moving is now explicit. Test it in the quiet weeks after Cyber Monday, judge it on 30 conversions, and see whether you get anything like 18%.

Microsoft may well have built a better bidder. The way to find out is to make sure you can see it working, and right now Microsoft has not given you that view. Build your own before Friday.