Tuesday Oct 30, 2012 by Larry Kim - Founder and CTO, WordStream
On the heels of Google’s “disappointing” earnings announcement last Friday – they generated a whopping $10.8 billion in advertising revenues in Q3 2012, up 5% from the previous quarter and up 16% year over year – I conducted an analysis to shed some light on just how they pulled this off.
The results are summarized in our Google Statistics Infographic, also shown below (click the image to enlarge). Read on for more commentary on what we discovered, including facts about Google advertising metrics and the top 10 industries that spend the most on Google ads, as well as some very detailed FAQ’s on how all this data was compiled!
Google is an advertising company – excluding the Motorola business, 94% of Google revenues come from advertising, which relies primarily on key advertising metrics like Impressions, Clicks, Cost Per Click, and Click-Through Rate.
In a nutshell: Google Ad Revenues = Ad Impressions * Click-Through Rate * Cost Per Click.
My analysis shows that, in the last quarter, the average costs per click (CPCs) have declined significantly in the last quarter (-16.5% for Google Search, -18.2% for Google Display Network), while click-through rates (CTR) were mixed (-12.4% for Google Search, +13.8% for Google Display Network).
Offsetting those declines are an impressive growth of ad impressions and clicks (clicks were up +21.6% for Google Search, +29.1% for Google Display Network), which more than made up for lost revenues.
It’s hard to say for certain what is causing what.
The dramatic increase in impressions and clicks is in some way contributing to the decline in average CPC’s and CTR’s – generally speaking, higher supply means lower prices, and showing a greater number of ads on a page inevitably means that any one individual ad is less likely to be clicked on. Alternatively, the massive increase in impressions could be a deliberate strategy on Google’s part to monetize more of their search inventory to increase clicks and revenues.
I think it’s probably a bit of both. In the end, the huge increases in impression volumes and clicks edged out the declines in cost per click and click through rates, resulting in yet another record quarter.
My take is that the trends we’re seeing in the Google economy create a win-win for both AdWords advertisers and Google.
A larger available inventory of impressions, combined with lower CPC, means that PPC advertisers are now literally able to get more customers for less money.
It also opens up Search Engine Marketing to more advertisers, including perhaps advertisers for whom the economics of search might not have previously worked out at higher average costs per click. (Just as SEO is harder for small businesses with lower budgets, so is PPC.)
Here at WordStream, we work with small and medium-sized businesses with limited search budgets ranging from $1k / month to around $100k / month. And previously, we’ve been quite concerned about rising CPCs and how it often results in smaller businesses responding by adopting increasingly narrow ad targeting parameters, or even dropping paid search all together.
I recognize that CPC’s aren’t fully controlled by Google per se – that an advertiser’s actual cost per click is a reflection of advertiser competition for a keyword, as well as an advertiser’s historical performance track record (Quality Score). But regardless, I’m very supportive of a Google Advertising system that, over time, emphasizes increases in ad inventory and provides features to improve click-through rates and conversion rates – because it delivers much more value to advertisers in the long run and makes paid search a much more sustainable and attractive venue for ad dollars, in comparison to other advertising venues.
Numerous other interesting findings came out of this analysis, including the following Google facts:
I used the data collected by the AdWords Performance Grader, a free AdWords account audit tool which has evaluated over $1 billion dollars in annualized spending on Google in the last year (roughly 2.5% of total advertising revenues on Google!). For this report, I used metrics from accounts that were evaluated between July 1, 2012 and October 30, 2012.
We believe this is the most comprehensive study on the internal working of the Google AdWords system, ever.
My research included over 2,600 AdWords accounts that ran our AdWords Grader in Q3 2012. The accounts in aggregate represented over $250 million in annualized spend. They ranged from very small (spending under $100 / month) to very large (millions per month in spend), across every industry. Additionally, the analysis includes accounts from all countries where Google does business.
I believe our data set is the broadest, most diversified cross-section of AdWords advertisers that is truly representative of the heterogeneous customer base that Google AdWords has today, and as such, is far more accurate than other studies by search marketing vendors that only work with say, just large clients.
I looked at 2,600 AdWords advertising accounts – examining key metrics like the total number of number of clicks, impressions, costs (etc.) or all of the accounts in the date range of July 1, 2012 to September 30, 2012. I first calculated some top-level metrics, specifically:
I also determined various important ratios, including the split of revenue and clicks split between Google search (including partner search) and the Google Display Network.
When determining conversion rates, I ignored data for accounts that didn’t have Google Conversion tracking enabled, which was roughly half the dataset.
The next step was to figure out key top-line stats (across all industries) for the entire Google AdWords platform, including:
I was able to determine these numbers using the Google Q3 2012 CPC’s (search & display) and Google’s total advertising revenues of $10.8 Billion, as well as the various ratios that I had previously calculated (share of revenue to Google Search, Google Display, Average CPC on Google Search, etc.).
The next step was to do the same calculation based on specific industry (e.g., Finance, Travel, Shopping, etc.).
To do this, I determined the various ratios and metrics (CTR, CPC, etc.) using advertiser data for a specific category, such as Travel. Next, I determined some industry weightings to apply. For example, what share of total clicks or total spend on Google belong to Travel, etc.
Once I had determined the industry weightings and the industry-specific metrics and ratios, I could then develop the metrics on an industry-specific basis.
Yes. I would like to disclose the following known issues pertaining to my research:
Currency Conversion: In my research, I converted various international currencies (for example, Euro, South African Rand, UK, Pound, Indian Rupee, Danish Krone, Australian Dollar, Singapore Dollar, New Zealand Dollar, Canadian Dollar, Norwegian Krone, Swiss Franc, Japanese Yen, etc.) using a fixed average conversion rate for the entire quarter. A more accurate way to do this would have been to use a different currency exchange rate for every day to account for daily changes in exchange rates.
Rolling Average: The dataset included in the analysis are accounts that were analyzed in Q3 2012. However each account analysis is a snapshot of the advertiser’s account performance metrics over the last 90 days. So, for example, an account evaluated on July 1, 2012, includes data for June 30, 2012 and the 90 days prior, which is actually Q2 2012. So what you’re looking at is a rolling average of data, not just data in the quarter.
High Conversion Rates: My analysis assumes that conversion tracking is properly set up by the advertiser, which is not always the case. A future data index will look for unrealistically high conversion rate outliers and throw those out. They were, however, included in this analysis, which resulted in higher conversion rates.
Higher Margin of Error for Industry-Specific Data: While the number of advertiser accounts used in the analysis was a very large number (2,600 for Q3 2012), the sample size becomes smaller when I break down the analysis based on the advertiser’s industry, which results in greater margins of error. I believe that the accuracy of our data can be further improved by addressing the various known issues and increasing the number of accounts analyzed per quarter.
Higher Weighting of English-Speaking Countries: Because our AdWords Grader is marketed to English-speaking markets, this likely resulted in a higher concentration of companies from English speaking-countries being included in the analysis. A future analysis will segment customer data for USA vs. Rest of World.
Rather than trying to make our research match up exactly to the every one of the few the bits of data that Google discloses, I chose to do a bottom-up analysis of the Google AdWords system using all of our available data, and fully disclose the research methodology.
I de-duplicated the accounts in our dataset. If an advertiser ran the AdWords Grader two or more times during the quarter, the older data from the previous run was discarded from our analysis.
To categorize the data into industries, I relied on an advertiser self-reporting their industry pick-list of 20 industries, prior to running the AdWords Grader application.
Couple of points here:
If you’ve read our research on the most expensive keywords, you might be thinking that the costs per click on AdWords are much higher. But actually those are the outliers in specific geographies with high competition.
CPC’s vary based on advertiser competition. For my analysis, I was looking at the entire Google AdWords platform, so I included data from advertisers worldwide, including countries with lower competition and lower CPC’s than the USA.
We found that CPC’s for individual ads were down -16.5% and -18.2%. We realize this is more than what Google reported in their earnings report – please see an earlier question on known issues for more information, particularly the “Rolling Average” issue. I believe what I am seeing is 2 quarters of declines in CPC’s (keep in mind that Google reported that their CPC’s fell in both Q2 and Q3 of this year).
We think it’s related to having more ads on a page. That if you have double the number of ads on a page, there are more things to click on, so naturally, the click-through rate of an average ad decreases.
We’re also not sure how exactly Google calculates their click-through rates – are they looking at the average click-through rate for a page containing multiple ads, or at the click-through rates of each individual ad?
Finally, keep in mind the “Rolling Average” issue – that we’re seeing 2 quarters worth of data, not just one.
I’m actively working on ways to improve our dataset – I hope to release regularly updated paid search industry benchmarks in the next quarter or two.
I’m very interested in hearing your thoughts on what should be included in the data index going forward, in addition to the metrics that I included in this report. For example, share of Mobile Search vs. Desktop Search, etc. Let me know in the comments below!
Larry Kim is the Founder & CTO of WordStream in Boston, MA. You can find this post, as well as additional content on WordStream's blog located here. You can also follow Larry on Twitter (@larrykim) by clicking here.