Your Costs Went Up and You Did Not Touch Anything
Some time in the second half of August, a lot of Google Ads accounts got quietly more expensive. Not because anyone raised a bid. Not because a competitor arrived. The cost per click on campaigns that had been running untouched for months started drifting upward, and the reports that used to show comfortable numbers started showing ordinary ones.
If that happened to you, the first instinct is to go looking for what you broke. That is the wrong search. What changed is that on August 17 Google switched off a mechanism that had been holding your bids down, and most advertisers never knew it was running in the first place.
We wrote about this change before it landed, when all anyone had was Google's announcement and a set of predictions. Five weeks later there is real data, including data from accounts we run. This is what actually happened, who it actually hit, and which of the decisions you were offered in August turned out to be the right one.
Google Switched Off a Mechanism It Never Told You Existed
Here is the part that reframes everything else.
When a campaign was capped by its daily budget and you had also given it an efficiency target, Google did not simply try harder within the cap. It lowered your bids. Deliberately. Google's own explanation of the behavior, surfaced by the team at Smarter Ecommerce, reads like this:
Read that second sentence again, because Google is describing a downside it built on purpose. This hurts your ability to compete fully in auctions.
The practitioner shorthand for it is bid suppression, and the side effect is the thing you probably liked. Lower bids meant you only ever won the cheap auctions. Cheap auctions meant cheap conversions. Cheap conversions meant a campaign that beat its own target month after month, which felt like the account being clever.
It was not the account being clever. It was a throttle. And the price of the throttle was every auction you were quietly not allowed to enter.
The mechanics of what replaced it, and the ten-minute check you should have run in your own account, are in the piece we published before the change. What follows is everything we could not know then.
Your Lost Impression Share Was Measuring Google's Throttle, Not Your Competition
Open any campaign report and you get two columns explaining why you did not show up more often: impression share lost to budget, and impression share lost to rank. Budget means you ran out of money. Rank means you were not competitive enough on bid or quality.
Every advertiser reads those two numbers the same way. Lost to budget is a funding problem. Lost to rank is a you problem.
Now look at what budget-limited campaigns were actually reporting before August 17. Smarter Ecommerce found the median budget-limited campaign was losing roughly 45 percent of its impression share to rank, and only about 4 percent to budget.
That is backwards, and it should have bothered more people than it did. These are campaigns Google itself labeled limited by budget. They should have been losing impressions to budget. Instead the overwhelming majority of the loss was landing in the rank column.
The reason is now obvious. Your bid is one of the main inputs to ad rank, alongside your quality signals. Suppress the bid and you suppress the rank along with it. So the system lowered your bids to protect your budget, and then reported the consequence as a rank failure, which reads in the interface as though your ads were simply not good enough.
Since the week of August 17, those two lines have crossed. Loss to rank fell from roughly 45 percent to roughly 30 percent. Loss to budget climbed from roughly 4 percent to roughly 33 percent. The label finally matches the mechanism.
There is a practical consequence worth sitting with. If you spent the last two years chasing a high lost-to-rank number on a budget-capped campaign, rewriting ads and pushing at quality score to fix something that was never a quality problem, that effort was aimed at a number Google was generating itself. Going forward the column means what it says, which makes it worth trusting again.
The Cheap Clicks Did Not Disappear. They Went to Advertisers With Room to Spend.
The suppressed bids were not creating cheap clicks out of nothing. They were parking budget-limited advertisers in the cheapest corner of the auction and leaving them there. That corner had a finite amount of inventory in it, and when the throttle came off, it got released.
Smarter Ecommerce measured both sides of that release. Among budget-limited campaigns that use a return target, median cost per click rose 15.8 percent after August 17. Over the same window, median cost per click on campaigns that were never budget-limited fell 13 percent.
Both halves of that sentence matter, and the second half is the one nobody is talking about.
If your campaigns are capped, you are paying more. If your campaigns have room to spend, you are paying less, because a pool of inexpensive inventory that used to be locked up by throttled advertisers is now available to you. Same auction, same day, opposite outcome, and the only thing separating the two groups is whether the campaign hits its daily ceiling.
That is an uncomfortable read for a small advertiser, so let us be plain about it. If you have been running a tight budget against an ambitious target, you were being subsidized by a mechanism you did not ask for and could not see. You are now paying closer to the real price of the traffic, and part of what you gave up is showing up as a discount in somebody else's account.
We Went and Looked at a Real Account
Industry medians are useful and easy to argue with. So we went into an ecommerce account we run and checked whether any of this survived contact with a real profit and loss statement.
The account had something close to a controlled experiment sitting inside it. Two shopping-led campaigns, both using a return target, running the same products to the same market in the same weeks. One was pinned against its daily budget, spending 95 percent of it in August and slightly over 100 percent by September. The other had real headroom and spent roughly three quarters of what it was allowed.
Only the first one meets Google's criteria for the change.
We compared two pooled windows, the first two weeks of August against the last week of August plus the first week of September, so the comparison is not just measuring a bigger budget. Campaign A was carrying a target of 5.3 times return, which Google Ads shows as a 530 percent target ROAS; Campaign B a target of 3.35 times, or 335 percent.
- Campaign A, pinned at its cap. Cost per click went from $1.25 to $1.41, up 13.1%. Return came down from 5.8 times spend to 4.7.
- Campaign B, running at roughly 75% of its cap. Cost per click went from $1.51 to $1.48, down 2.4%. Return came down from 3.8 times spend to 3.0.
Look at the cost per click first, because that is the clean result. The capped campaign paid 13% more per click. The campaign with headroom paid slightly less. Two campaigns, same account, same products, same weeks, opposite directions, and the only thing distinguishing them is whether they hit their daily ceiling. Smarter Ecommerce measured the same split at a much larger scale and got a wider spread, up 15.8% against down 13%. We got the same shape at a smaller magnitude from data we pulled ourselves.
Now look at the return figures, because they are the more honest and more uncomfortable finding.
Both campaigns' return came down by about a fifth. Not just the exposed one. The campaign that does not meet Google's criteria for this change, the one that should have been unaffected, moved by almost exactly the same proportion.
That matters enormously, and it is the reason we are not going to tell you this change wrecked anybody's return on ad spend. If August 17 were driving the decline, the capped campaign should have dropped and the campaign with headroom should have held. It did not hold. Something else in that account, whether seasonality, product mix, a promotion cycle or the conversion timing problem we come to in a moment, pulled both campaigns down together.
So our own data confirms the cost per click story and refuses to confirm the return story. If we had only looked at the capped campaign, we would have written a much more dramatic and much less true article.
Your Account Average Will Hide This Completely
Here is the finding we did not expect, and it is the most useful thing in this article.
At the account level, that same account's cost per click held flat, moving 1 percent. One. If we had opened the account summary, glanced at the trend, and moved on, we would have concluded that August 17 did nothing at all.
The campaign inside it moved 13 percent.
We then ran the same comparison across twelve accounts we manage that had enough volume to say anything. The result was genuinely mixed. Some accounts showed cost per click up by half again. Others were flat. Several were meaningfully down. There was no consistent pattern across the book, and no reliable relationship between how budget-constrained an account looked and which direction its costs went.
We could dress that up as a study. It is not one. Twelve small accounts, each with its own campaign edits, seasonal swings and budget changes over the same six weeks, cannot isolate a platform effect. The honest conclusion is that at small-business budget scale, the noise from your own decisions is louder than the change.
Which is exactly why the advice is not to go looking at your account average. It is to do this:
Filter your campaign list to the campaigns showing the red limited-by-budget status. Of those, keep the ones using a target cost per acquisition or a target return. That is your exposed set, and it is usually much smaller than people expect. Compare each one's cost per click and its actual performance against target across early August and mid September. If a campaign is not on that list, August 17 did not touch it, and whatever else is moving in that campaign has a different cause you should go find.
Anyone Telling You What It Did to Your Return Is Reading Unfinished Data
There is one more thing standing between you and a clean answer, and it is the reason we are being careful with the return figures above rather than leading with the scariest number we found.
Conversions do not all arrive on the day of the click. Someone clicks on a Tuesday, thinks about it, comes back the following week. Your most recent weeks are always missing conversions that have not landed yet, which makes recent performance look worse than it is going to turn out.
We measured how bad that gap was in the account above, rather than assuming. The most recent full week we pulled was only 66 percent attributed. A third of the value belonging to that week had not arrived at the time we looked. About 12 percent of all value in the account lands after the first day, with a tail stretching out past a month.
So every alarming chart of collapsing return on ad spend circulating right now, including the ones we could have drawn from our own most recent week, is partly an artifact of looking too early. We dropped that week entirely. Smarter Ecommerce, who have far more data than we do, said the same thing in different words: the cost per click and impression share story is measurable today, and the return verdict is not in until conversion data finishes attributing toward the end of September.
This is the opposite side of a trap we flagged in August. Then, the risk was setting a new target off an unfinished 28 days and accidentally making it stricter than anything your campaign had ever hit. Now, the risk is diagnosing the change off unfinished weeks and panicking at a number that is going to move. Same incomplete data, two different ways to get burned by it.
If you want a real answer on what this did to your return, the date to run that report is the end of September, not today.
Which of Your Four Options Aged Well
Back in August the choice was between keeping your target, lowering it to match recent performance, setting a number your margins actually support, dropping the target altogether, or funding the campaign properly. At the time those were five reasonable options and we could not tell you which one the evidence favored, because there was no evidence yet.
There is now, and it points fairly clearly at one of them.
Raising the budget got better, and it got better for a reason that did not exist in August. Non-limited campaigns are the ones seeing cost per click fall 13 percent. Moving a campaign out of the capped group does not just buy you more volume anymore. It moves you into the group being charged less per click for the same traffic. That is a genuinely different calculation than the one available before the change, and it is the option almost nobody took. Smarter Ecommerce noted only a small minority of advertisers went this way.
Lowering your target to match recent performance aged the worst. Two reasons. First, the conversion lag problem above means your recent actual is probably flattered, so you will set it too tight. Second, our own two campaigns each came down by about a fifth on figures that are not finished attributing, which means the number you would be anchoring to is softer than the one you will end up with.
Keeping your target is still correct if the number is genuinely what a customer is worth to you. Just go in knowing what you are buying: performance drifts toward that number, your volume falls, and your reports say you hit your goal the entire way down.
Dropping the target for a maximize strategy still does what it says. A campaign with no target cannot be budget-limited in the sense this change cares about, because there is no efficiency number for the system to aim at.
And the question underneath all four is still the one worth asking. If a campaign has lived permanently in the red limited-by-budget state, that was never a setting to optimize around. It was a decision nobody finished making. We worked through how to tell whether your budget is genuinely below your market's floor in Google Has No Minimum Spend. Your Market Does., and the durable version of how targets behave, independent of any one announcement, is in Your Google Ads Bid Target Isn't a Wish. It's an Instruction.
What August 17 really did was stop Google from making that decision quietly on your behalf. The bill for an underfunded campaign used to arrive as a smaller, cheaper campaign than you realized you had. Now it arrives as a higher cost per click, in a column you can actually see.
That is a worse-feeling month and a better-run account. If you would rather have someone watching which of your campaigns are sitting in that exposed set, and deciding on purpose rather than finding out in a report, that is what we do. Either way, the campaigns to go look at are the capped ones, and the date to judge them is the end of the month.
