SEMrush: From Where Does the Data Cometh?

  

 

From Where Does the Data Cometh?

Although I have been using SEMrush for a couple of years now, I still feel like I have barely scratched the surface of its seemingly endless and amazingly powerful capabilities. Much like Google Analytics or Google Ads, each time I log in to perform a task it seems that there is a new tool available, often teased with an icon indicating “upgrade today to access” or something to that effect. Rather than find that burdensome, I find it refreshing as well as reassuring to know that large teams of developers are constantly pushing the capabilities envelope. And yet, with each new feature, each new report, each new anything, I still have the realization that exactly NONE of it would work without a vast, clean and robust data set to be mined. Each time I open SEMrush I literally do ask myself, “so where are they getting all this data exactly?”

To find the answer to this question I began to research SEMrush’s data acquisition techniques on their website. Their blog states, “SEMrush uses its own machine learning algorithms and trusted data providers to present the data in our databases. There are different methods for gathering different types of information but the high standard of quality across our databases remains the same. We only use the most up-to-date data sources, and we always clean the data through our proprietary methods in order to present the most trusted solution on the market.” It then continues, “For search engine rankings and keyword analytics, we use third party data providers to collect Google’s actual search results pages for the 500 million most popular keywords. Then, we collect information about the websites that are listed in the top 100 positions. We study both organic search results as well as paid search results to give you a complete picture of any website’s visibility on Google” (SEMrush, n.d.). 

So, if I’m reading this correctly, it sounds to me like SEMrush is using third-party vendors to run search queries on Google, then they’re somehow ingesting all of that information, sending it over to SEMrush who then runs data-cleansing scripts on that data, before making it available to customers. Although that sounds like a mind-boggling amount of work and data, it does make sense because I can’t imagine that Google just hands over an API key to anyone and says, “here’s our data, pay your license fee and you can just have at it.” Not happening. 

Then my question becomes, “since this is data that has been captured programmatically, then been put through several rounds of cleaning, how accurate is this data?” Doing some research, I came across a number of articles that seemed to align on the key takeaway that, “SEMrush data is far from perfect, but it is the best tool out there.” One author went so far as to compare his Google Analytics traffic with that of SEMrush, and found discrepancies stating, “It’s hard to measure traffic when you’re not Google and you don’t have access to the site. But like I said, my goal is not to pick holes in SEMRush but to remind website owners that this is not the accurate tool they think it is. Nothing is. It’s a guideline and it should be taken as such. Most webmasters know that, but there are those who take it as the most credible source” (Editor, 2018).

I have actually performed similar comparisons on several client sites, as well as my own sites. In each case, I did find that the exact metrics were typically somewhat off, but the relative metrics were essentially very accurate and reliable. In other words, if SEMrush had a pageview count that was 10% off what I was seeing in Google Analytics, it had the same variance everywhere making the relative comparisons accurate. And I would put forth the question, “how accurate is Google Analytics anyway?” Over the past few years, I have seen more and more bots and bogus traffic get into my Google Analytics reports, and skew data sets to the point where I’ve gone into client meetings wondering if I was going to have a bulletproof answer for anything. And if Google—with deeper pockets and more seasoned talent than just about any other company on earth—hasn’t figured all of these things out, well then in my opinion no one can. 

For keyword research, the SEO team at Siperia.com conducted their own tests and found, “We performed Google searches for 20 keywords and matched the SERP results against the organic keyword rankings presented by SEMrush. Out of 20 keywords searched, the SERP results in google.com were only mismatched to SEMrush’s keyword report 4 times. And for every ranking that did not correlate, the difference in rankings was only 1 position, indicating the accuracy of SEMrush’s keyword data” (Siperia, n.d.).

The point is that even if SEMrush is operating on a slightly (or even more than slightly) imperfect data set, it is still by far the pre-eminent platform in this space. Further, I have used this tool in practice to find actionable insights that I’ve presented to clients, with which they have made critical business decisions and come out ahead. That to my mind is the only validation I need in practice, and I do sleep easy at night knowing that the development team at SEMrush is likely quite aware of their data shortfalls and are working both diligently and passionately to continually improve. As I stated previously, each time I go into the tool I find something new and improved “above the hood” which makes me believe they are similarly committed “under the hood” as well.  

References

Where does Semrush data come from? (n.d.) SEMrush Blog. Retrieved from: https://www.semrush.com/kb/998-where-does-semrush-data-come-from

Editor. (17 May 2018). Why SEMRush Data Is Not as Accurate as You Think. Home Business Magazine. Retrieved from: https://homebusinessmag.com/businesses/ecommerce/search-engine-optimization/semrush-data-not-accurate-think/

SEMrush Review. (n.d.) Retrieved from: https://seperia.com/blog/competitive-intelligence-tool-reviews/semrush/