SEO PROJECT

The odd ratio shows the relative probability of belonging to a group compered to any other lower ranking group.

In the case of traffic for example 1 percent incresement with everything else staying the same increases the probability of belonging to a higher category with 46 percent. Based on the p values the traffic seems the most significant variable, which paired with a higher OR than expected implies an important conclusion. As the traffic and the ranking are in a reciprocal relationship with each other, we can guess that keeping a good position is easier than reaching one.

By the other significant value, the social media I expected monotone decrisment from the first to the fifth level, so that the more social media sites are a page present, the higher the probability is for reaching a better position. In contrast, my results showed the highest value in the fifth and second category.

The other variables are not proven significant, so I didn’t analyse their odds ratio. Based on the p value the increasing of the variables (which in the case of the categorical variables means turning to 1 from 0) doesn’t effect the reaching of a better position.

​

​

​

My results differ from the findings of previous research on several points. Nagpal and Petersen's (2020) study looked at rankings as a function of online authority, content relevance, search progression and degree of competition. In their model, online authority was included as a combination of page authority and domain authority, which gives a reason to compare the results with the backlink variable in my model, which is also related to the value of page authority. Their results show that, unlike mine, the variable is significant at the five percent significance level. The difference, beyond the fact that the indicator in their case contains information not only about the URL but also about the whole website, may also stem from the fact that their variable is relative, with values relative to competitors. I did not consider this kind of relativisation to be important due to the similar search volume between the variables considered in my research. In addition, the two researchers considered the first fifty results, which may also cause reach, due to the claim described earlier that the algorithm may not necessarily use the same ranking among the top search results as it does on either of the other pages of the search result.

​

Also for comparison, the Content Relevance variable from Nagpal and Petersen's (2020) research, for which the title of the pages was examined, provides a comparison. Their observation is significantly different from my result for the variable cim, which is significant at the 5% significance level. In the case of my database, this variable already showed the weakest relationship with the outcome variable when the Cramer indicators were calculated and was not found to be significant in the model. This discrepancy may also be due to the difference in data collection, as the two researchers also used a more sophisticated method to investigate the semantic relationship between the keyword and the other terms in the title.
Also in contrast to my results is the research of An and Jung (2019), who also found that the optimization of meta fields can improve the ranking of a URL when designing an appropriate keyword targeting method. In contrast, the title, meta description, and meta keyword were not found to be significant in the model I built.

​

​In terms of the frequency of use of these tools, however, my finding, mentioned in the descriptive statistics section, is in line with the work of Leun and Chan (2021), who also found that the most frequent of the on-page tools is the mention of a keyword in the title, followed by a small lag in the frequency of mentioning it in the meta description, and the two indicators are followed by a much smaller proportion of mentioning the keyword in the meta keyword.