In an earlier post, it shown that the relationship between CTR and position can vary for different advertisers, especially in the higher positions. These differences can potentially be explained by differences in brand traffic or are real differences in the click-through related to the dynamics of geographical PPC markets or the nature of the advertiser’s business. It is interesting to note that the CTR in the higher positions seem to be considerably better for advertisers in less mature online markets namely Australia and South Africa. We investigated this further by only focusing on non-brand specific keywords. To achieve this we included a factor indicating the brand or non-brand status of keywords in our model. Figure 1 compares the four advertisers on all non-brand keywords with a QS of 7. Even after non-brand differentiation we continue to observe substantially higher CTR in the top positions for Australian and South African advertisers compared to their UK & US counterparts used in the same model. This could be reflective of a lower level of competition relative to more developed markets resulting in a smaller number of competitive ads on a search results page, which in turn drives higher click-through in top positions. From this it is also clear that there are no fixed thresholds for determining quality score, rather it is determined on the relative CTR which varies across advertisers and potentially also geographical regions
Figure 1: CTR by position across four advertisers for all non-brand keywords with a quality score of 7.
The factors that affect quality score are hotly debated in many blogs and online forums. This motivated us to look for answers in our own data. From what we see in our own data, we have little doubt that click-through rate (CTR) is the most important factor in determining quality score (QS). When studying the relationship between QS and CTR it is important to take into account the effect of position in that analysis which is often over-looked in other analyses. Clicks2Customers, as a company, works with clients in a wide variety of industries and PPC markets around the world, which allowed us to make comparisons from the above dynamics across four clients from PPC markets in different geographical regions.
Logistic regression is a useful and appropriate statistical tool for studying the relationship between a binomial variable or rate, such as CTR and other factors (in our case quality score and position). We selected 4 clients representing a range of industries and geographical locations and modeled the observed CTR as a function QS and position. It is important to note that QS does not affect CTR (in fact causality happens conversely); however the regression model is a useful descriptive tool for highlighting the underlying relationship between the above factors.
In figure 1 we plotted the above relationship model for the four selected clients. Similar to the analysis referred to above, it was observed that quality scores of 8 and 9 are rare, therefore focus was on quality scores of 6, 7 and 10 where there is more data to model. Furthermore, we restricted our analysis to the top positions. The relationship between CTR and both QS and Position proved to be significant, though perhaps predictable. Clearly, a higher quality score implies a higher CTR at a given position. It is clear that the QS for a keyword is determined by the CTR relative to the position the keyword is at. For example, let us consider the data for the large US retailer in figure 1. Keywords in position 1 with a QS of 10 have a CTR of around 10%, while a QS of 10 corresponds to a CTR of around 5% in position 3. In case you are concerned about the use of CTR in determining QS related to position, you have no concerns as it seems as though Google has thought it through.