Click here to read more on what Google has say about  the quality score.

For more information about the ad action you can watch this very useful video.

So I was very curious to find out if what Google communicates in blogs, websites,and customer support is true and how strong type of relationship between the keyword CTR and keyword quality score.  So I took 1677 keywords from various campaigns that play to Google and Google Search Partners 1677 keywords from various accounts and various business activities. I have include some keyowrds with low search volume but I have excluded all the keywords with 0 impressions and 0 clicks. What is interesting is to examine the realtionship in the healthy and working keywords So I created 2 data sets.

• Data set 1 includes all the keywords with statistics from at least the past 2 years.
• Data set 2 includes the same keywords as dataset 1 with statistics from the last 2 months (10/11/2009 -10/01/2010 )

Let’s see some descriptive statistics from data Set 1 in order to understand what kind of keywords we are dealing with.

Correlation Coefficient R

Before making a simple analysis let’s talk about Correlation Coefficient R. The correlation coefficient indicates the strength of a linear relationship between two variables.
The correlation coefficient, denoted by R, is a measure of the strength of the straight-line or linear relationship between two variables. The correlation coefficient takes on values ranging between +1 and -1. The following points are the accepted guidelines for interpreting the correlation coefficient:

• 0 indicates no linear relationship.
• +1 indicates a perfect positive linear relationship: as one variable increases in its values, the other variable also increases in its values via an exact linear rule.
• -1 indicates a perfect negative linear relationship: as one variable increases in its values, the other variable decreases in its values via an exact linear rule.
• Values between 0 and 0.3 (0 and -0.3) indicate a weak positive (negative) linear relationship via a shaky linear rule.
• Values between 0.3 and 0.7 (-0.3 and -0.7) indicate a moderate positive (negative) linear relationship via a fuzzy-firm linear rule.
• Values between 0.7 and 1.0 (-0.7 and -1.0) indicate a strong positive (negative) linear relationship via a firm linear rule.
• The value of r squared is typically taken as “the percent of variation in one variable explained by the other variable,” or “the percent of variation shared between the two variables.”

Want to learn more?

Wikipedia article http://en.wikipedia.org/wiki/Correlation_and_dependence
http://www.dmstat1.com/res/TheCorrelationCoefficientDefined.html
http://hsc.uwe.ac.uk/dataanalysis/quantInfAssPear.asp

CTR and Quality Score relationship

Data Set 1 (All time)

Data Set 2 (last 2 months)

In both cases the correlation coefficient  R  is very high ( R1 = 0.445 or R2 = 0.367) which means a very strong relationship between quality score and CTR that is also  statistically significant (pass the test for 99%).That definitely means that keywords CTR is a very strong indicator of quality score

The interesting bit has to do with the fact that R is much bigger 21% (R1 = 0.445 VS R2 = 0.367) for data Set 1 (all Period). This means that Google takes seriously into account not only the last 2 months of data but also  the historical CTR of this keyword. If you take into account the fact that  adwords is interested in the historical data of ads texts/ ad group / account then history plays a very important role in adwords performance (success). We cannot expect big accounts to have excellent results from day 1. A short term solution for this is to have an increased budget and bigger maximum CPC especially in the first running weeks. This will gives us a good quality score in the beginning and we can try to maintain that. If we begin with a low quality score then the only solution is to open a new account and start from the beginning.

Let's see some scatter plots for both data sets:

Data set 1 All period

Data set 2 -Last 2 months

As we had expected, there’s a positive linear relationship between Quality Score and CTR.

Correlations

Let's find the correlations between Quality score and other keywords variables

The colors are the variables that are statistical significant ( 99% confidence ). Pearson Correlation is the R. Taking a closer look the most important positive relationship is CTR ( R=0.445 ) and follows a negative relationship with first page CPC (R= - 0.323 ) and some less significant negative relationships for Impressions and Average_Position (R = -0.194 ) . This means that Google takes into account the Average position and thus utilizing it two ways. Firstly the average position affects Quality score directly and a low average position affects CTR.

Can we build a model for the quality score ?

The simple answer is no. Building a reliable and robust model would need much more that the variables we have in our hand. We need to quantify the quality of the landing page, to quantify the relevance of the ad with keyword, search query and landing page, to quantify the bad/good history of the account/ad group and many others variables. However let's play with the keywords variables to find a simple linear model (with linear regression). I will use the data set 1 because it contains historical data and we could build a more accurate model.

In plain terms this model says that CTR is very good for the quality score and first page CPC is not. This is very logical because keywords that have a high first page bid are very competitive and expensive and most of the time have lower CTR than the rest because of the difficulty to achieve a good position.

For the this model I used linear regresion with STEPWISE method and variables (First_Page_CPC Impressions Clicks CTR Avg_CPC Cost Avg_Position). The R is 0.456 ( R square 0.208 and Adjusted R Square 0.207 )

Model 2 for the quality score

Quality score= 7.093 + 9.053*CTR - 1.303*FIrst_page_CPC + 0.296*Average_CPC

The interesting part here is that quality score is improving with Average CPC. I have read in many cases that Google incorporates in the adwords algorithm that ads in low positions cannot have high CTR so it does not penilize them. This model shows that this is partial not truth. You can have higher quality score if you have higher average CPC and better position.

For model 2 I used linear regression with ENTER  method and variables First_Page_CPC CTR Avg_CPC Avg_Position. R is 0.532 ( R square 0.283 and Adjusted R Square 0.281 ) and statistical significant with 99% confidence.

Main Findings

1. CTR is the most important factor of the quality score
2. Having high First Page CPC is negative for the quality score
3. Google needs to otpimize it's revenues so it tells you to increase the CPC if you want to increase Quality score :)
4. High average position is not good for the Quality score

Strategy - Actionable Insights

This a big discussion but I think that there are 2 strategies

Strategy 1 - I want to be the king of my keywords. Top positions many more clicks

• I want to have high average position (1-3 ) in the majority of the keywords
• I use all keywords long tail + head keywords
• I have a website and can handle the cost and can monetize the high cost traffic
• I can pay higher than the competitors for the same keywords ( especially in the beginning)
• We optimize continuously adword account

Strategy 2- I can live in lower positions

• I care a lot about the CPC even in the begining of the account
• I can live with a small percentage of the traffic from my keywords
• My low bids gives me low average position and low CTR
• I have a low share of relevant traffic (from PPC) but I can live with that
• I care more about ROI than in absolute conversions and absolute revenues/profit
• I know that if I want Strategy1 I need to be patient or I have to change my adwords account
• I do not spend much time/resources in optimizing campaigns

I think these strategies are the extremes and there are all sort of variations between them.You can choose whenever fits you as long you know the advantages and disadvantages. My personal opinion is that I want to start with strategy 1 because it is better to do the right thing the first time.

Do you have any comments? What do you think would be interesting to analyze? What is personal opinion about the 2 strategies?

Any personal opinions about the variables of quality score and the success of the adwords?