Why data science?
Using data science allows professional SEOs to get the statistical edge in SEO by helping to make better decisions on what aspects of web design, content and offsite SEO are beneficial and detrimental for SEO. Data science has the advantage over humans in that the methods are scientific, i.e. the findings need to be robustly tested including double blind tests in order to ensure the discoveries are significant and the results are repeatable. Using data data science to predictively model search engine algorithms such as Penguin and Panda also means that coincidences and ‘bad thinking’ are much less likely to bias or destroy the value of the discovery from the data that can often happen as humans often take shortcuts when interpreting data. The benfits by using data science are as follows:
- Evaluate offsite links for disavowing or link removal
- Getting “buy in” from clients to make onsite changes
- Rapid recovery from Search Engine Penalties
- Understanding which links are powering the top ranked competitors
So how is data science better than the SEO tools currently available?
The SEO tools of today can’t deal with artificial intelligence
As search engines like Google use artificial intelligence by deploying machine learning into algorithms like Panda which uses Support Vector Machine (SVM), there is an increasing need for SEOs to build or make use of data science platforms. The truth of the matter at the time of writing is that SEO’s simply don’t have the tools or the skillsets to cope with the pace of change of the search engines. Whilst we may have lots of access to data, SEO experts aren’t exactly known for their statistical modelling skills. The industry leading figures still quote impressive sounding names such as “mean spearman correlation studies” even though they are constantly qualifying themselves with “correlation doesn’t equal causation”. Even worse, the link API products that are currently available in the market are great for modelling PageRank. However none of them can provide diagnostic value for Panda or Penguin or Hummingbird because the latter algorithms are completely different and therefore require completely separate models.
Why MathSight is a game changer
MathSight is a game changer because the methods of data science are employed to statistically model the search engine algorithms. MathSight’s team are comprised of mathematical experts that have significant experience including the industries of Formula 1, Aerospace & Defence, Finance/Banking and Oil & Gas. The mathematical techniques borrowed from these industries are applied to the data set we collect from sites they crawl and have site analytics access to. When MathSight crawls these sites, they look for signals much in the same way a search engine does. Google claim to only process 200 signals whilst Yandex claims 800 signals. MathSight also crawls and extracts signals from the webpages that make up the link graph for the data site’s inbound links from external sites.
Building models to discover causal relationships
Using the signals extracted, the MathSight platform will process the data to mathematically identify patterns where signals are common to pages that win or lose traffic. Of course such patterns may be a coincidence which is the consistent disadvantage of correlation studies. MathSight then uses a wide variety of probabilistic models to see which of those signals is significant (i.e. 95% or more) in causing a loss or gain in SEO traffic. Only then is a signal seen as significant and the signal’s benchmark can be considered reliable for helping SEO’s optimise for.