ClickEquations Blog
A Weblog on Paid Search Marketing, Search Analytics, and Online Marketing
Secret Truth Series #13: The Average Lie
Numbers look like facts even when they’re not.
To make the best possible PPC management decisions it’s important to know the difference between the numbers in your reports which represent something that has happened and those that are mearly clues that need more investigation before you can possibly know what they really mean.
PPC Reports Don’t Present Raw Data
With thousands of keywords and hundreds of thousands of clicks (or more) in each account every day, we should be glad that the search engines or our management platforms don’t try to show us the raw data behind our accounts. It would be too much to handle.
Instead what we see are lots of sums and averages.
- The number of clicks on a keyword is the sum of all clicks from each individual search query that was matched to that keyword.
- The number of clicks on any search query is the sum of all clicks from each geography or person.
- CPC, position, revenue-per-click, and many others are averages – sometimes labeled as such and othertimes not. We’re never told the distribution of the data.
Look Behind The Sums
The first place we look when reviewing our paid search accounts is at campaign and ad group level performance data. Here every number is a sum or an average.
These reports can be very useful when we’ve come to know and trust the data behind these numbers. But it’s easy to mistakenly assume or presume that the summed numbers tell you something they don’t about the contents of any one group.
- You can’t see when two data elements cancel each other out. For example, if one ad group surged by 50% while another fell by 50%, the campaign data can report a flat month-over-month number. You can miss important changes ‘inside’ the data.
. - You can’t see when one data element overwhelms the others. If an ad group shows a terrible ROI or CTR, there is no way to know that one or more keywords inside are hugely profitable or have fantastic CTRs. The good can overwhelm the bad, or visa versa.
When looking at campaign, ad group, and even keyword reports, of course you want to react to the numbers, watch their trends, and compare them to their peers. But make sure you know the numbers behind these numbers.
Before taking any action, dive in and look at the numbers underneath. You may not always have the time to make the necessary changes – so you may still need to take the broad action – but you may find that in the long run the change you had planned is not the right long term solution.
Don’t Believe The Averages
When AdWords tell us that a keyword had an average position of 3, it’s hard not to think and act as if it was ‘usually’ in position 3, or even that it was almost always at or near position 3.
But this may or may not be the case.
As Google Analytics shows, most keywords (their ads really) get placed in a very wide range of different positions and the average reported in the interface is just that – an average. The keyword you think about as having been in position 3 was probably seen by many people while in position 5 and seen by others in position 1.
Click To Zoom
This keyword was in positions as diverse as Top 1 and Left 6 during a 1 month period.
Averages without standard deviations – as someone used to tell me – are rather worthless.
The joke he told to illustrate was about a comedian told to prepare material for a birthday party where the average guest was 35 years old. As you might imagine, he included a little colorful language in the routine he had planned. Only to be shocked to arrive and find a birthday party filled with 3 year olds and their 73 year old grandparents.
Taking the distribution of the data that makes up the average into account is especially difficult when you don’t have access to it.
- We see quality score reported for our keywords, without any way to find out the different quality scores earned across the different geographies where the associated ads were shown, or the impact of the different text ads that the keyword displayed.
. - We see CPC data for each keyword, without any way to find out the different CPCs that were actually charged for each different search query that was matched, or the impact of the quality score, geography, ad copy, etc. (Did you even know that CPC was an average too?)
Don’t Believe What You Read
The Boomtown Rats had it right. The numbers we get back in PPC reporting systems are not always what they appear. Teach yourself to view them a little skeptically, and definitely not conclusively. Analysis always has to follow data.
Sums and averages are necessary because we can’t usually handle raw data. My point isn’t that these reports aren’t useful or presented properly. But rather that it’s important to know what the data is and what it isn’t, and to know how to take the next step to learn more about any particular number that raises questions.
In many cases the numbers behind the numbers are just a click away. You just need the time and inclination to go look at them. In other cases, the engines still aren’t sharing enough details, and we all need to pressure the engines (with whom we spend so much money) to provide us with the data we need to make fully informed decisions.
What Do You Think?
This blog post is part of a series extending and amplifying the ideas in our free ebook ’21 Secret Truths of High-Resolution PPC’.
What they’re saying: “If your goal is to be the best you can be at paid search, then your path goes through this book. When Craig talks I listen, mesmerized. You should too because being wise is great!.”
- Avinash KaushikAuthor of Web Analytics 2.0 ’.
.
Secret Truth Series #11 – How AdWords Quality Score Impacts CPC
The Max CPC and quality score of a keyword determine its position, and position and quality score drive actual CPC. So exactly what effect does quality score have on cost?
We first answered this question one year ago, in the now famous ‘Economics of Quality Score‘ post. (This has since become the most visited page in the history of this blog.) You should go read this now if you haven’t already.
The central chart from that article shows the percentage discount or penalty you pay for every click based on your quality score.
If you know how many of your keywords are receiving each quality score, and the amount of your spend on each, it’s easy to use this data to calculate the total cost of poor quality scores, savings from great quality scores, and the net cost to your account.
Incidentally, ClickEquations provides this as a default report – isn’t that handy?
This Is Probably True
The only caveat to these calculations is the little-known-fact that quality score IS NOT a number between 1 and 10.
Google reports quality score to us mere mortal advertisers using that scale, but in the great AdWords super-computer a wider range of values is used – so your actual quality score may be 37 or even 68.2394.
We don’t know the range of numbers they use, the number of digits of precision, nor the relationship of one score to another.
And while this isn’t a secret truth, the fact is that I’m not much of a mathematician. So at the risk of public scrutiny and embarrasement, here’s the logic that lead to the above quality score impact calculations – feel free to issue corrections and admonishments in the comments:
CPC is calculated by dividing the ad rank of the advertiser below you by the quality score of the advertising keyword. The table was created by calculating the difference between dividing X by 7 and dividing X by 8. This difference, it turns out, is consistent regardless of what X is equal to.
Therefore, if quality scores were really whole numbers between 1 and 10, the chart above should be accurate.
Since they’re not, we don’t know (at least I don’t) the impact of a different range of quality score numbers which act as divisors. If a perfect quality score is really 83 and not 10, and a very good quality score is really 64 and not 9, there would be a difference in the percentage impact to CPC of earning a perfect quality score versus and very good one.
The assumption made in publishing these numbers as they are (which was disclosed) is that the real levels are proportionally similar. That could be wrong. Which means that the discounts and penalties on the extremes could be more or less. There is no way – short of a Google announcement – of knowing.
I believe the numbers to be directionally true. Perhaps as Jim Sterne said about web analytics in general, they’re ‘true but not accurate’.
What Is True
The details probably don’t matter anyway. Quality score does in fact apply as a discount or a penalty to your CPCs. And whatever the numbers, the farther your quality score is from the mean, the more severe its impact.
What matters is that we realize that high quality scores save us money (and get your ads shown more frequently and in higher positions) and that low quality scores cost us money (and result in less ad display and lower positions). In terms of data, everything after that are merely interesting.
In terms of action, we need to use that knowledge to drive our actions. We want to be aware of our keyword quality scores, and manage them, based on the fact that they drive our placement and to a very large degree our costs.
What Do You Think?
What they’re saying: “The glory of paid search is hyper relevance and how absolutely data driven it is. If your goal is to be the best you can be at paid search, then your path goes through this book. When Craig talks I listen, mesmerized. You should too because being wise is great.”
- Avinash Kaushik Best-Selling Author ’.
.
Match-Type Rock Scissors Paper
NOTE: This is part of a post series. It’s available as a single post for easier reading: The Match Type Series.
===============================================
In several earlier posts in this series I’ve discussed the how’s and why’s of buying the same or similar terms at the same time with different Match Type settings.
I outlined in one post the details of creating a Match Type Keyword Trap to filter certain search queries into specific match types. Buying multiple terms and multiple levels – when done correctly – has the ability to give you control over which queries are caught at which price.
Exact-Phrase-Broad
One reason this works is because the engines (generally) execute the match types sequentially.
In other words, if you are bidding on the same keyword, or two keywords that would both match for one particular query, an Exact Match should take precedence over a Phrase Match which should take precedence over a Broad Match.
So even though a particular query is technical a match for both one Broad Match keyword and another Phrase Match keyword, the Phrase Match should always ‘win’ and catch that query.
I should hasten to point out, this will not always be true. If you carefully watch query reports for your keywords you will see queries that were exact matches against a keyword you had set to Exact Match, yet the query lands in a Broad Match group. But in our experience these are rare in the sub 1% range of all queries.
Emphasis the Match Type Setting with Higher Bids
You can and should add punch to this precedence by ALWAYS placing rather substantially higher bids on your Exact Match vs Phrase Match, and Phrase Match vs Broad Match when they’re stacked in targeting the same terms.
And make the differences between the bids significant – it generally won’t help to bid $0.05 more for Exact Match than Broad Match. When bidding it’s easy to look at your Max CPCs (since that’s the option used to set the bid) but since your actual and average CPC is usually just a fraction of the Max you really can’t base your decision on those. Look instead at average CPC’s being reported and then set the Max’s at large enough intervals to create real steps between the different keyword/match type combinations.
By placing a substantially higher bid on the match type differentiated keywords, you’re providing another algorithmic reason for the engine to match exact match queries to your Exact Match keywords. Of course, it should also be true that you want generally higher position and higher impression share for the keywords you’re bidding on Exact Match.
A Simple Match
At the start of this series I mentioned that Match Type was a powerful and often under-utilized option. I hope these five posts so far have covered some of the ways you can get more out of these options. Time for a break from Match Type, however. Watch for a new series starting soon.
Facebook
Twitter
YouTube
Blog
Comment or Question? Leave It Here!