Rank tracking using a single method does not provide accurate results. There are several factors involved in a search such as personalization, localization, data centers, inclusion of not provided, etc., which can affect the process of locating the true ranking of a keyword.
A study was done using data of Moz.com and four methods were used to determine ranking. These four methods were Google Chrome Personalized browser, Chrome Incognito browser, Crawler, and Google Webmaster Tools. 500 queries were analyzed for a month to track click through rates and rankings.
All forms of personalization were removed as best possible by using &pws=0 in the MozCast engine while logging out of accounts when using browsers. In Google Webmaster Tools, the parameters were set to Web and United States. The result was 206 queries that contained full data.
Correlations were found to be strong on both Personalized vs Incognito in browsers and rankings were found to be grouped quite closely on higher spots. There was lower correlation on Crawler vs Google Webmaster Tool.
The 4 methods were also compared to look for agreement or disagreement between the methods in terms of correlations. The results showed that Google Webmaster Tools (GWT) had the most disagreement with the other 3 methods and Incognito showed the least disagreement.
Further study showed that personalization had a big impact on a few queries, but that effect was only limited to a few searches. Personalization also seemed to be penalized in a few cases. However, this aspect is quite complex to fully comprehend. On the other hand, GWT results contradicted other findings by a noticeable margin.
While numbers provided by GWT may be right, there is no way of learning how they are deduced by Google. Google’s reasoning seems to be that SEOs should merely trust them to provide rankings without offering details on how they measure and provide results.
At the end of the study, it seemed that all 4 methods used did not differ too much from one another. Although each method had its limitations and faults, the results were somewhat realistic. However, these methods do not include local results, do not have crucial SERP features such as Knowledge Graph and Answer Boxes, and do not provide information on click through and traffic.
Future studies do need to consider an all-round view with regards to visibility and rankings, irrespective of the methods chosen to measure it. You too should use more than a single method to get a clearer picture while tracking rankings.