How Google is changing the PPC Auctions to hit targets

Let’s start with a high-level perspective: Google’s main driver for revenue is the Google Ads business. They have to deliver growth. How can you achieve that when the number of users stays the same or, even worse, when you start losing traffic to alternatives like ChatGPT Search?

Why am I writing this post now?

There were two recent changes in Google Ads that made me really suspicious:

• Google removes Auction Insights from Looker Studio: Starting from Sept. 23, 2024, Looker Studio reports will no longer show data from the Auction Insight Reports. Yes, the CSV data is still available in the Google Ads Report Editor. However, it’s not that convenient anymore and a lot of advertisers will be blind after the change. As a solution, we recently shared a free auction insights tool.
• Google experiments with double ad placements for the same advertiser (Posted on Dec. 18, 2024, in Search Engine Land): Anthony Higman (ADSQUIRE) observed multiple ads of the same advertiser and was wondering how this “double serving” is possible. Ginny Marvin (Google Ads) commented it’s due to a “small experiment”. It’s unclear since when the experiment is running. I guess some weeks already.
  

More competition = more ads revenue for Google

There are two levers to increase competition:

1. Increase the total number of Google Ads Auctions — even if the number of users searching on Google is constant.
2. Increase the number of Auction members across all running auctions.

How to model competition: Merge Daily Auction Insight data on keyword level with Keyword Performance Data

Yes, the data gets quite big, and you won’t handle the data in Excel or Google Sheets on bigger accounts. Here is in short what I did:

Daily Keyword Data, Keys for joining the Auction Insights data: Day, Keyword ID, Ad group ID:

Daily Auction Insight Data, Keys: Day, Keyword ID, Ad group ID.

The following transformation was made: I grouped the data by day, keyword ID, ad group ID and counted the unique “Display URL domain” items. These are the competitor domains you are familiar with in auction insight reports. What I get: after that transformation, I get the exact number of auction players on keyword level for every day of the year. This new dimension was joined with the keyword performance report:

My money formula (if I worked for Google Ads): Maximize sum(#Auction Players * Impressions)

To be fair: Only looking into a single number can fool you when there are some seasonal effects in the industry. For that reason, I added the total number of impressions as baseline on a second axis. When the curves spread apart, something is changing.

So what is happening here?

• First: the data is coming from a single account targeting the US only. The keyword set and the bidding strategies haven’t changed over time. The sample size is reasonable (yearly spending> $1 million)
• Quarter 1: the blue line (Impressions) was always above the red line
• Mid-March: both lines are on top of each other now, this means competition was growing relative to the impression baseline. Explanation: that’s the high season within that industry, high season stops from one day to another mid-April. After that, the blue line is above the red line, like at the beginning of the year.
• Starting in July 2024: the competition (red line) is growing relative to the impressions (blue line)
• Mid of August 2024: The Growth rate is increasing week by week
• Mid of September 2024: The new normal is reached: there is way more competition in every auction now
  

And what does it mean for the average CPC?

It’s interesting: in the highest seasonality in April, the CPC was stable. With the recent changes in the auction competition, the CPC curve is perfectly following the “#Auctionplayers * Impressions” — Curve from above.

How is this possible? My guess: Close variant matching

There are some ways to measure that, all the following charts show the same time period as the figure from above.

String Similarity of Keywords vs Matched Queries

The shown chart is the string similarity (Levenshtein) of the keyword and the search term that was triggered. The lower the value, the lower the similarity.

Starting in August, the similarity was going down slightly. Another drop happened around mid of September.

Search Term Vocabulary (normalized with Impressions)

What does it mean: Based on the search terms that are available in the reports, we can count the vocabulary of words that appeared. If the number of words increases, it’s probably due to changes in the matching logic: when you are more fuzzy in the matching, you’ll end up having more vocabulary in the search term report. The data is normalized with impressions (it’s normal to have a higher vocabulary count in higher traffic seasons — that’s considered with the normalization).

The chart tells the same story as the string similarity.

Cost Share of Reported Search Terms

I know: the hidden search terms are an old topic already. Google says this all happens due to privacy regulations. Advertisers are suspicious about it and think that Google is hiding information.

What should I say? The Cost Share of queries that are available in the search term performance report is going down, matching the movements of the charts addressing the matching behavior from above. Interesting.

Summary, and how can you deal with it?

1. Google is hiding, once again, data that helps advertisers to understand what’s happening in the Google Ads system.
2. The competition level has been heavily forced since September 2024. No surprise when the quarterly Google numbers will go up.
3. Less of the total costs can be explained with known search terms that appear in Google reports. However: That’s still one big action area to combat what we are seeing here. Use negative keywords heavily for off-topic queries/search patterns!
4. If you run any bidding strategy that is based on impression share: Review that right away! The impression share will go down right away if there are more players in the auction. Google will pay higher bids for you when the impression share drops because of that.

Final thoughts

I’m personally waiting for the tipping point of Google’s Ads business. When Google is losing traffic shares to new alternatives, the only way of hitting the revenue goals is to further squeeze out the advertisers. Until they can’t afford it anymore.