Somewhere in your inbox or your Slack channel right now there is a sentence that goes something like this: "Our return on ad spend has been down since August, and it's the Google update."
It is an easy sentence to believe. Google really did change how its target bidding works on August 17, the change really does push some campaigns toward a lower return, and the timing lines up neatly with a lot of soft Septembers. When a number drops and a platform change landed the same month, the platform change is the obvious suspect.
Obvious is not the same as guilty. We run an ecommerce account where exactly this explanation came up, and when we pulled two full months of data apart campaign by campaign, the August 17 change turned out to be a small part of the story. The bigger pieces were sitting in the account the whole time, and none of them get fixed by blaming Google.
This article is the follow-up we promised in September, now that the conversion data has had time to finish arriving. It covers what the industry has actually measured, what our own account showed once the data matured, and five checks that will tell you whether your decline really is August 17 or something you can do something about.
Two Months After Google's August 17 Change, Nobody Has Published Its ROAS Impact
Start with what is known, because it is less than the volume of commentary suggests.
The cost side has been measured. Smarter Ecommerce, a bidding software company that watches a large number of shopping accounts, published data in mid-September showing that the median cost per click on budget-capped campaigns using a return target rose 15.8 percent after the change, while campaigns that were never capped saw their median cost per click fall 13 percent. That is a real, measured effect, and it is the one we walked through in Google Took Your Cheap Clicks and Handed Them to Someone Else.
The return side has not been measured, at least not publicly. Smarter Ecommerce said the return verdict would have to wait until conversions finished attributing toward the end of September. As of mid-October, that follow-up has not appeared. Search Engine Land, Search Engine Roundtable and the bidding tool vendors have published explainers and early cost data, but no return figures across a set of accounts. PPC Land summed up the state of play on October 10 in one line: "Evidence on the effect remains anecdotal."
Google has also quietly updated its own help page. It now says the rollout of the change "completed on August 27, 2026," ten days after the date everyone remembers, and recommends waiting one to two conversion cycles after a change like this before making account decisions. For most stores that means a month or two. In other words, anyone who diagnosed this change in the last week of August was looking at a change that had not finished arriving, measured with conversions that had not finished landing.
So when someone tells you the update cut your return, the honest response is that nobody, including Google, has shown that at scale yet. It might be true for your account. You have to check.
Google's August 17 Change Is Real. It Only Squeezes Campaigns Pinned to Their Daily Budget.
None of this means the people complaining are imagining it. Some of the reports are genuinely ugly.
Grow My Ads, an agency with more than a hundred clients, described one campaign set to a 400 percent target that had been delivering 600 to 700 percent before the change. Its share of shopping impressions fell from a peak of 69 percent to under 10 percent, and sales were down more than 35 percent. On the PPC subreddit, one advertiser with a shopping campaign that used to return 30 to 40 times its spend reported that "sales dropped significantly," and another said shopping campaigns that ran above 10 times their spend now sit around 2.2. Hold on to that second one; it comes back in Check 3.
Notice what those campaigns have in common. Every one of them was capped by its daily budget and was beating its target by a wide margin. That is exactly the campaign Google said would change: when a budget-limited campaign has a return target, Google used to hold its bids down to stretch the budget, which produced a return well above the target as a side effect. Since the change, it bids up to the target to spend the budget instead, so the return drifts down toward the number in the box. We explained the mechanics before the change landed in On August 17, Google Starts Taking Your Bid Target Literally, and the more a campaign was overshooting, the further it had to fall.
Grow My Ads makes one claim that goes beyond Google's description, and it is worth taking seriously. They say budget-capped campaigns with no target at all, running Maximize Conversion Value, also lost impression share, and so did a capped campaign that was already landing on its target. If they are right, the change reaches further than Google said. But even their wider claim stops at the same place: the accounts they say were untouched were the ones that were not capped by budget.
That gives you the one fact every diagnosis has to start from. The August 17 change works through the budget cap. A campaign with room in its budget has nothing for this change to take away.
Our Budget-Capped Ecommerce Campaign Kept Its ROAS Through August 17
In September we compared two campaigns in the same ecommerce account: one capped by its budget, which meets Google's criteria, and one with room to spend, which does not. Both had lost about a fifth of their return, which told us the change was not the main driver. We also said the return question would not be settled until the conversion data matured. It has now, so we ran it again with two full months after the change and a cleaner set of windows.
Here is how each campaign's return moved, using the date each sale happened so recent weeks are not undercounted.
Campaign A, capped by its budget (the one the change applies to):
- July 1 to August 5: 6.4 times spend, leaving out three unusual days covered in Check 2
- August 6 to 16: 5.8 times
- August 17 to September 2: 5.7 times
- September: 4.9 times
- October 1 to 15: 4.8 times
Campaign B, room in its budget (the change does not apply):
- July 1 to August 5: 3.8 times spend
- August 6 to 16: 3.6 times
- August 17 to September 2: 3.0 times
- September: 3.6 times
- October 1 to 15: 3.0 times
Read the capped campaign first, because it is the one the change can reach. Across August 17 itself it barely moved, from 5.8 to 5.7 times spend. It did step down by about 13 percent from early September, and its cost per click climbed from $1.24 in mid-August to $1.44 in September while its impression share rose. Higher prices and more impressions on a capped campaign is the change's known pattern, so a delayed effect there is plausible. It is also the same week that campaign's target was edited, which comes up in Check 3, so we cannot hand all of it to Google.
Now the campaign with room to spend. It took the bigger fall of the two, right at August 17, from 3.6 to 3.0 times spend, while its cost per click stayed flat around $1.48. The change cannot touch this campaign. Whatever moved it, it was not August 17.
So the account did lose return over these two months. The change explains a modest slice of one campaign. The rest came from the five things below, and every one of them is something you can check in your own account in an afternoon.
Check 1: A Campaign That Never Runs Out of Daily Budget Is Untouched by August 17
This is the sixty-second check, and it settles more arguments than any of the others.
Open your campaign list and look at the status column. You are looking for campaigns marked limited by budget, and for each one, whether it actually loses impressions to budget. In the columns menu, add search impression share lost to budget. A campaign that loses a meaningful share of its impressions to budget is capped. A campaign losing close to zero is not, whatever the status label says.
In our account, the campaign that carries most of the spend lost under 3 percent of its impressions to budget in every week from June to October, and usually close to zero. Google's change has no lever on it. If your biggest decline is in a campaign like that, you can stop reading about August 17 and start looking for the real cause.
There is one thing an uncapped campaign does that often gets mistaken for the update. A campaign with a return target and room to spend will spend until its return settles near that target. That is not new, and it is not the change. If your target is set at 3 times spend and the campaign used to deliver 3.8 on a smaller budget, then as the budget grows, the return walking down toward 3 is the campaign doing exactly what the target tells it to. We wrote the longer version of that in Your Google Ads Bid Target Isn't a Wish. It's an Instruction. If that return is no longer good enough for the business, the fix is the target, not a complaint about Google.
Check 2: Three Big Days in July Are Not the ROAS Benchmark You Lost
Every comparison has a "before," and most people never look at what is inside theirs.
Our capped campaign returned 9.1 times its spend from July 1 to August 5. Measured against that, everything after looks like a collapse. But three days in that window, July 20, 21 and 24, brought in $44,725 of sales between them, on about $1,200 of spend. That looks like a handful of large orders landing in the same week. Take those three days out and the same five weeks returned 6.4 times spend. The "drop" from July to late August shrinks from more than a third to about 12 percent, before the change has even entered the picture.
This happens constantly in stores that sell anything expensive. A $2,000 machine or a bulk order from a contractor is not something a campaign produces on a schedule. When three of them land in the same month, that month becomes the standard everyone measures against, and every normal month after it looks like a failure.
Here is how to check your own baseline. In the campaign's report, set the segment to day and sort by conversion value, highest first. If the top two or three days carry a large share of the period's revenue, run the comparison again without them. Then compare a typical day, not the total: put each period's daily return in a list and take the middle value. Our capped campaign's typical day was about 5.5 times spend in both July and mid-August, and 5.2 in late August. The July average was carrying three outliers; the typical day barely moved.
Check 3: The ROAS Target You Raised That Week Moved Your Results Too
The most common edit around this change was not made by Google. It was made by advertisers.
Smarter Ecommerce tracked what happened to the targets people set. On budget-capped campaigns, the median return target rose 4 percent the week before the change and 12 percent the week it landed. That number normally barely moves. More surprising, campaigns the change did not touch at all, the ones with room in their budget, saw their median target rise 9 percent the same week, and overall spend across their data fell about 10 percent. A lot of advertisers got nervous and tightened their targets on campaigns that had nothing to fear.
A tighter return target means Google skips auctions it would have entered. That can cut volume, shift which products get shown, and change your return in either direction. If that edit happened in your account the same month as the update, your before-and-after comparison is measuring both at once.
The Reddit advertiser from earlier, whose campaigns fell from above 10 times spend to 2.2, is a clean example. By their own description, they raised those campaigns' targets on August 17 itself, splitting the difference between the old target and what the campaigns had been delivering. Two changes landed on the same day, and only one of them was Google's.
Our account had two edits of its own in the same window. On August 6, eleven days before the change, the capped campaign's daily budget went from $390 to $551, a planned push for more volume. Its return started stepping down that same week, and its cost per click rose from $1.10 to $1.24 before August 17 arrived. Then on September 3, its return target was raised slightly, the same week its return dipped again. A budget increase that size buys deeper, pricier auctions on its own, change or no change.
To check yours, open the change history for each campaign and filter it to budget and bid strategy changes from July onward. Put every edit on the same timeline as your return. If someone in your business, your agency, or a Google rep touched the target or the budget near the change, test that first. It is the one cause you fully control.
Check 4: Our ROAS Dipped Just as Far Last September, With No Google Change
Before you compare this month to last month, compare it to the same month last year.
In our account, return fell 7 percent from August to September in 2025, with no Google change anywhere near it. In 2026 it fell 8 percent over the same two months. The September dip everyone is blaming on the update is, in this account, almost exactly the size of last year's ordinary September dip. For a lot of stores, the end of summer simply sells less per dollar.
October is where this year actually breaks from last year. In 2025, October bounced back above August. In the first half of October 2026, return fell further, and average order value fell to about $211, its lowest of any stretch since July. That is a real problem worth chasing, and it is a much more useful finding than "it's the update," because it points at what customers are buying rather than how Google is bidding. A store selling fewer big-ticket items in October needs a different fix than a store paying more per click.
The check is simple. In your reporting, set the date range to the period you are worried about and add a comparison to the same dates in the previous year. If last year had the same dip, you are looking at a season. If last year did not, keep digging. For a sense of how much a market's own rhythm moves the numbers, our piece on the minimum your market really costs walks through why those swings are bigger than most owners expect.
Check 5: New Campaigns and Three-Cent Clicks Distort Your Account's ROAS and Conversion Rate
Your account's overall return is an average of everything running in it, and averages hide things in both directions.
Two things in our account were pulling the average down that had nothing to do with August 17. The first was a test. From mid-September to early October, a new Search campaign using Google's AI Max features spent about $5,700 and returned 0.67 times its spend before it was switched off on schedule. New campaigns in a test period are supposed to underperform while they learn, but its spend still sits inside the account's monthly figure.
The second was stranger. In the last two weeks of September, one small promotional campaign picked up about 18,000 clicks at around three cents each. Its total cost was under $600, so it barely moved the return. What it did move was everything that counts clicks. The account's conversion rate appeared to fall from 2.4 percent in August to 1.6 percent in September. Take those clicks out and September's conversion rate was 2.4 percent, essentially unchanged. Anyone reading the account summary would have concluded that shoppers stopped buying. They didn't. A cheap, low-quality stream of clicks diluted the number. That campaign now needs its own cleanup, but it is a separate problem from the bidding change.
The check: before you diagnose the account, look at it campaign by campaign. Sort by spend and ask, for every campaign that started, paused, or changed budget since July, how much of the change in the account total it accounts for. Then scan for any campaign whose cost per click is wildly out of line with the rest. A campaign paying a few cents a click is usually buying low-quality placements, and it will distort every rate in the account.
If August 17 Really Did Hit Your Campaign, Lifting Its Budget Cap Ends the Squeeze
Run all five checks and you may still have a capped campaign that was beating its target, kept its settings, and lost return and volume from late August onward. That one is the change, and it is worth treating differently from the rest.
Agencies dealing with it are testing a short list of moves, and none of them is a guaranteed fix yet:
- Remove the target for a few weeks. Switch the campaign to Maximize Conversion Value without a target, let it settle, then set a new target from what it actually delivers. Grow My Ads is doing this on its worst-hit campaigns.
- Start a fresh copy. Some agencies are duplicating the campaign under a new ID and launching it without a target. Nobody has shown this works better than the first option; it is an experiment.
- Cap the expensive clicks. A portfolio bid strategy, even on a single campaign, lets you set a maximum bid. Set it at the level where your most expensive clicks stop producing sales, not at your average cost per click, or you will cut off auctions you need. Google's own FAQ on the change advises against adding bid limits purely in reaction to it, because they can cause swings, so save this for a campaign that is clearly overpaying for clicks that never sell.
- Fund the campaign. A campaign that is no longer capped is out of reach of this change entirely, and since August it also joins the group whose cost per click went down. If the campaign is profitable at its target, more budget turns that into volume.
The full set of options, and how to pick the right target from your margins rather than your dashboard, is in our August walkthrough. One caution that applies to all of them: make one change, then wait one to two conversion cycles before judging it, which is what Google itself now recommends. A string of quick edits creates the same tangled before-and-after this article is trying to untangle.
Blaming Google's August 17 Change Is Not a Strategy. Five Checks Take One Afternoon.
Here is the order to work in. Find which campaigns are actually capped. Take the big days out of your baseline. Line your own edits up against the dates. Compare to last year. Then look campaign by campaign for the new and the strange. Whatever is left after those five checks, if anything, belongs to August 17.
In our account, those checks turned "the update did it" into a much more useful list: a target doing what it was told on the biggest campaign, a July baseline inflated by three big days, a budget increase made the week before the change, a seasonal dip the size of last year's, a test campaign and a stream of junk clicks. Every item on that list has a fix. "Google did it" doesn't.
Blaming the platform is comfortable, because it means nothing needs to change on your side. It is also usually wrong, and it costs you the months you could have spent fixing the thing that actually moved. If you'd rather have someone run these checks on your account and manage the bidding from here, that is the work we do every day. If you'd rather run them yourself, everything you need is above, and it takes an afternoon.
