Google’s August 17 Bidding Update: Good or Bad? The Data

Episode Transcript

00:00:01
Chris:
LFG! GM!

00:00:05
Mike:
Welcome to another episode of Growing Ecommerce. We’re starting to get into the Halloween spirit with a very spooky autopsy: a postmortem of the August 17th Smart Bidding update. Chris, what are you laughing at?

Chris: It’s the best intro! I’m so pumped now. I love it.

Mike: I’m pumped too. You know what? Halloween is a big thing in our house.

00:00:29
Chris:
Yes, same here.

Mike: It doesn’t get much scarier than the August 17th Smart Bidding update. Holy smokes. We promised weeks back that we would follow up on this. You were getting messages, I was getting messages—people saying they appreciated the last episode and wondering where the heck the new episode was.

00:00:51
Mike:
It was a big episode because it seems like a lot of people got moved by it emotionally.

Chris: Factually. Factually moved.

Mike: Maybe that is a good segue to the facts we talk about today, because there are super interesting insights to discuss.

Chris: Talking about Halloween horror stories, by the way, the next episode will be Halloween.

00:01:17
Mike:
Yeah, we’re going to have a Halloween bonus episode next week.

Chris: Awesome! Talk about horror ads and e-commerce horror stories.

Mike: Yes, because the horrific part of that story is, once again, the communication of that bidding algo change, which was just absolutely horrific and brought a lot of uncertainty.

00:01:41
Chris:
What can really be boiled down to that blunder is that it was badly communicated. Because at the same time, let’s face it: from the numbers that can be shown today, there might be a case to be made that this move made a lot of sense. That’s my intro. Maybe I hand over to you to guide us through this story.

00:02:03
Mike:
Yeah. I mean, it still feels ambiguous to me. Was this good or bad? Should we be enthusiastic about this or not? I think a lot of people are rightly upset. And definitely, while Google improved the communication as they went, it was not a strong point of this launch.

Chris: Maybe for potentially new listeners, a short wrap-up of what this whole thing was about, and then we dig into it.

00:02:26
Mike:
Sure. So basically in late June, I believe, buried at the bottom of a blog post, Google made an announcement that there would be updates to Smart Bidding in mid-August—when the whole Northern Hemisphere is on vacation. You were lying on the beach, right?

Chris: Yeah, I was on the beach!

00:02:45
Mike:
At any rate, this change affected limited-by-budget campaigns. To recap really quickly—because I think a lot of you are probably familiar with it—their problem (and we will fact-check some of this) is that too many of these limited-by-budget campaigns have superficially strong performance.

00:03:17
Mike:
You were asking for a ROAS of, let’s say, 5 and getting 7, 10, whatever, and Google wanted to change that situation. Because, to quote you, Chris: “Google owes you shit.”

Chris: Well, that’s the truth, yeah!

Mike: You’re getting something that you’re not asking for and not paying for. That’s part of it. At the auction level, this probably was some kind of hygiene act.

00:03:47
Mike:
I think it makes sense in many ways, but they wanted to resolve this situation. We’ll get into the specifics of how exactly it worked and what happened, but in a nutshell, they gave you a few options:

1. You can stop being limited by budget and spend more money.
2. You could increase your ROAS target to be closer to what you’ve been getting.

00:04:11
Mike:
3. You could change your bidding strategy to something else, like Maximize Conversions or Maximize Conversion Value, because these strategies cannot be budget limited.
4. Or you could do nothing and accept the new rules—accept what happens.

Chris: All right. And maybe just to make this a comprehensive wrap-up of what this whole thing was about: the hypothesis or goals behind that move on Google’s side was that they want to make sure there’s more ROAS accuracy.

00:04:47
Chris:
Right, exactly. Which should be the basis for every online retailer and everyone doing Google Ads—to make it easier for you to scale more smoothly because you precisely know if you put the ROAS at X, it will deliver around X. That was basically the whole thing.

Mike: Exactly. Yeah, they said more predictable performance, smoother scaling.

00:05:13
Mike:
I think we can evaluate the first half of that claim, but the second half is harder to evaluate, at least from today’s standpoint.

Chris: Because, again, we talked about that, and we will talk about it, and we’ll probably have some new numbers with us. But that is basically a good thing as an online shop owner: “Hell yes! More accuracy, great!” But there is a potential downside, and I think we have one or two use cases with us which precisely show what the downside actually can be. So yeah, hell of a topic, Mike. What numbers do you want to start with?

00:05:36
Mike:
Yeah, let’s evaluate this initial situation with some of our numbers. It was very hard to get down to what Google was actually saying in the end. In general terms, they mean too many people are getting a high ROAS and they’re going to change that, so your actual ROAS will trend towards your target.

Chris: Target, yes.

Mike: But what wasn’t clear was why it was occurring, and as a result, exactly how. We were speculating in the past: is that because your conversion rate is going to drop?

00:06:10
Mike:
Is your order value going to drop? Are CPCs going to be higher? What’s going to happen?

Probably the most useful slide out of everything I saw from Google—I looked at a lot of their materials—pretty precisely says what was going on. What they say there is that when a campaign is budget constrained, the system automatically decreases bids to avoid hitting the budget cap.

00:06:39
Mike:
They’re afraid that they’re going to spend too much money and have to refund you that money, so they throttle the bids on purpose. That’s what’s going on: the budgeting system lowers bids to stretch your budget out, leading to cheaper conversions, and this is what caused the surplus efficiency.

Chris: Yes. Just to add that perspective, theoretically, lower bids mean Google is focusing more on search queries that show the highest buying intent possible.

00:07:51
Chris:
In theory, the higher you bid, the more generic queries you will be exposed to—maybe upper-funnel search queries where buying intent is not so clear. Is that theory still true? And if yes, throttling bids should actually lead to higher conversion rates compared to higher bids. Is that still somewhat true?

Mike: It’s hard to say. Definitely…

00:08:19
Mike:
Back in the day, we did some analyses of CPCs. This was back in the days of search term sculpting. If you’re not familiar with that, you’d sculpt Shopping campaigns to target generic queries versus product long-tail and product-specific queries, and all kinds of different strategies there.

00:08:41
Mike:
What we found is that often on long-tail campaigns, where there’s a lot of intent, CPCs were actually cheaper, oddly enough, and you would often pay more on generic terms. I haven’t done an analysis like that in a long time, but what has always been the case with Google is that bids are very stratified—meaning there are layers in there, and you can end up hitting ceilings in certain kinds of intent and kinds of users.

00:09:14
Mike:
This creates situations where you run into local maximums, and then you can’t just linearly scale a little bit more. You suddenly have to increase your bids a lot to find a new level of growth pockets. Now, with Google’s bidding and targeting, I just don’t know.

Chris: Got it. However, the bottom line is they throttled the bids, but the value of those throttled bids was still there.

00:09:36
Chris:
And that’s why a lot of campaigns limited by budget created significant surplus. These were not bad, cheap clicks.

Mike: Right. What you actually want! And you could argue that these limited-by-budget campaigns were acting as a sponge—a market-level sponge soaking up good, cheap clicks.

00:10:05
Mike:
Now, the question is—and we have numbers for this—is Google right that these limited-by-budget campaigns were creating this surplus? The answer is yes, absolutely. Let’s show some numbers here.

Looking in general at how Target ROAS delivers, we’ve actually looked at over-delivery or overachieving targets in the past, which is handy because we had an idea about how this works.

00:10:37
Mike:
If you look at all Smart Bidding campaigns, what you’ll find is that when you have a small amount of monthly conversions, it’s very hit-or-miss whether the algorithm is going to be on target, above target, or below target. That can vary by day, week, or month. Certainly, the more conversions you have, the more data you can feed the algorithm.

00:11:02
Mike:
What’s very fascinating—and viewers can see this on screen—is a curve showing that as conversions increase, the percentage of campaigns underachieving the ROAS target diminishes, and the amount on target increases. But what’s fascinating is that the percentage of campaigns that over-deliver efficiency doesn’t change at all.

Chris: Exactly. So about a third of campaigns over-deliver, which is a lot!

00:11:42
Mike:
So that share is there, and we can confirm what Google sees when we look at the general population. What we learn is that if Google is going to miss the target, they try to miss it upwards rather than below. They know people react better to a ROAS that is too high than too low, even if you’ve theoretically left revenue on the table.

00:12:01
Mike:
Now, if we look at limited-by-budget campaigns and how they compare to that, instead of having about a third of campaigns overachieve, it’s 55%!

Chris: Holy smokes.

Mike: And only about a third are actually on target. So this is the problem Google is talking about, and we can check that claim off as true. Systemically, there is a problem caused by the mechanism Google explains.

00:12:29
Chris:
We can evaluate that claim as well, that Google is throttling bids which causes campaigns to over-deliver. If I as an online shop have a campaign that is limited by budget, and 50% of my limited-by-budget campaigns are giving double the ROAS I actually want—assuming the target ROAS I put on my campaigns is literally what I want to achieve—that makes a lot of sense, right?

00:12:54
Chris:
What do you do with all these products that are significantly overachieving the ROAS you are willing to accept? It means there is a lot of untapped market out there.

Mike: For sure.

Chris: So Google’s claim that they want to help you scale at an efficiency level that suits your business model—that’s the biggest claim I now fully understand.

Mike: Yeah, I think it’s healthy advice that you could spend more here because there is market out there at the ROAS level you are willing to accept.

Chris: Fair point. Shout-out to Google! Though we would have communicated it better, Mike.

Mike: Yes, maybe.

00:14:43
Mike:
So we established that, yes, these campaigns were over-performing. To evaluate the other claim that this was happening because of bids, I want to look at one of the big changes we saw on August 17th. It’s very interesting and confirms what Google is saying about bid throttling.

00:15:06
Mike:
Every search network campaign in Google has Impression Share, telling you how many of the eligible impressions you actually receive. There are two diagnostic metrics for Impression Share loss reasons: budget and Ad Rank. Ad Rank includes ad quality, relevance, landing page, and the bid. The bid is a crucial part of Ad Rank. If you’re losing Impression Share due to rank, it’s often because of your bid.

00:15:32
Mike:
Five years ago, landing page and ad quality massively dictated Ad Rank. Nowadays, the bid is by far the biggest indicator for rank, especially with Shopping and Performance Max. What’s really fascinating is that prior to August 17th, a very low percentage of these budget-limited campaigns were losing Impression Share due to budget—only about 5% or so.

00:16:22
Mike:
If you look at loss to rank, that was actually the dominant cause. Prior to August 17th, limited-by-budget campaigns were primarily losing impressions due to rank, not budget. After August 17th, those two lines cross. This shows Google stopped rank-limiting these campaigns to prevent overspending. Now they are genuinely budget-limited. Seeing those lines flip reflects reality much more effectively.

00:17:25
Chris:
This metric is actually a useful diagnostic now!

Mike: So we can check off a couple of claims from Google: yes, these campaigns were over-delivering efficiency, and yes, it was due to the bids. But what happened when they made the change? What happened to CPCs?

00:18:02
Mike:
As you might expect, bids went up and cost per click went up. Interestingly, there was a divergent trend: on budget-limited campaigns, CPCs went up by a pretty decent percentage, whereas they actually got a bit cheaper on campaigns not limited by budget.

00:18:58
Chris:
For campaigns limited by budget, CPCs went up because Google said, “Hold my beer—these cheap, very good clicks are no longer reserved for you at that target.” To bring effective ROAS down, they increased the bids. That’s one of their levers, so it’s not surprising. But what’s fascinating is why CPCs went down for campaigns NOT limited by budget.

Mike: I have two hypotheses. I love both of them—they could both be true, or both wrong, but only Google knows.

00:19:49
Mike:
Hypothesis one: budget-limited campaigns were acting as a sponge for cheap yet valuable clicks. That sponge got squeezed by Google, so those clicks went out into the market at large and got picked up by campaigns that could spend more, dragging average CPCs down.

Chris: I love that hypothesis. I think it makes the auctions more effective.

00:20:13
Mike:
Hypothesis two is just as strong, but way more rooted in irrational behavior. Right after these changes occurred, we looked at advertiser behavior. We found that not only did a lot of advertisers limited by budget increase their ROAS target (which makes sense as an option), but so did advertisers NOT limited by budget—which was completely unnecessary because the change didn’t affect them!

00:21:15
Chris:
Due to lackluster communication, people didn’t know what was going on, and out of insecurity, they increased targets for campaigns not affected by the algorithm by any means. This means there is market out there now: a more restrictive ROAS target from your competitor means clicks are available for you. For campaigns NOT limited by budget, don’t raise your ROAS target—actually be more aggressive now because there’s market up for grabs!

00:22:08
Mike:
A lot of advertisers increased their ROAS target because they were nervous about the change, driven by Google’s communication and influencers on LinkedIn, Twitter, Reddit, and YouTube. They turned their targets up and left them there. It is safe now to go back to what your target was before if your campaign was not limited.

Chris: Relax it to levels that make sense for you, and look into your market space—because if others are acting irrationally, there is market up for grabs.

00:23:45
Mike:
Let’s look at what happened to ROAS. Looking at how much the median campaign overachieved: back in June, limited-by-budget campaigns were about 20% above target. Non-limited campaigns overachieved by a much thinner margin—a median of 5%. Then there’s a cut, and behavior changes immediately.

00:25:09
Mike:
Within the week of August 17th, it drops down to about 11%. By August 24th, it’s about 6%. By August 31st, it’s at zero. So within three weeks, it hit target.

Chris: Didn’t Google claim there would be a smooth transition? I don’t know what “smooth” means in the Google universe, but if you drop from a ROAS level of 25 down to 10 in three weeks, that is NOT smooth!

00:26:04
Mike:
Everyone was worried August 17th would be a cliff. I put messages out there saying, “It’s not a cliff, Google says it will trend towards your target.” But actually, the data shows that it WAS a cliff.

Chris: So you were wrong, Mike! You were wrong! Am I a bad person to feel some schadenfreude telling you that you were wrong?

Mike: No, let me hear it! Say it as many times as you need to. It’s recorded—play it back to yourself falling asleep at night.

00:27:31
Mike:
Before this change, 55% of limited-by-budget campaigns were overachieving. Now it’s a much more even split: about 50% on target, a quarter above, and a quarter below. Google achieved what they set out to do, though not smoothly. That brings us to a specific case study we brought with us.

00:28:58
Mike:
This is a UK retailer in kitchen and homeware. There are two charts: actual versus target ROAS from July through early October, and weekly average CPC.

Chris: You can very clearly see the week of August 17th on the CPC chart. Before, CPCs were pretty cheap—around £0.17. Afterward, it immediately jumped up to £0.27 in one week—about 44% higher CPCs! That was a cliff, Mike! You were wrong!

00:30:47
Mike:
Shout-out to Google—it’s consistent with what they wanted to do, and they executed on it. Looking at the ROAS chart, in the before period, actual ROAS was super spiky and substantially higher than target—spiking up to 35, other weeks down around 10. Afterward, those two lines are married to each other. There’s no variance whatsoever.

00:32:09
Mike:
The client increased their target ROAS slightly to around 13, and actual ROAS immediately adhered to that line. That is predictable performance!

Chris: But there’s a massive downside. Retailers got used to this over-delivery and relied on it to meet business goals. Dropping from a ROAS 30-40% above target down to target in 2 to 3 weeks is serious. That’s my main criticism of Google: it was not smooth.

00:33:00
Mike:
Last night I did a webinar with Search Engine Land, and someone asked why Google is punishing them for being limited by budget. People definitely feel punished.

00:34:23
Mike:
Whether it feels like punishment or bitter medicine needed for auction health, this is the new normal. We need to accept where we are and move forward to have the best Q4 possible. Target ROAS is now a more reliable steering mechanism because actual ROAS tracks target closely.

00:36:15
Chris:
Think even more about ROAS as the primary steering mechanism to scale or be more restrictive. And there is a window of opportunity because the market is irrational!

Mike: Markets are always perfectly rational! That’s another episode of Growing Ecommerce. Brought to you by Smarter Ecommerce. Visit smarter-ecommerce.com to learn more. See you next time!