Should we toss out the whole GSC tool?I can understand why people say that GSC isn’t accurate, concluding the same things after much of my own research and analysis. GSC does have some discrepancies bringing its reliability into question. Should we then toss out the whole tool? GSC does provide strategic insights if one knows how to use the data. GSC landing page results line up reasonably well with Google Analytics (GA) landing page results as I will show later. Some accuracies can be improved upon by understanding how Google is collecting and showing the data sets. A key factor is understanding exactly what Google is showing, when you are slicing & dicing the data sets. There are also some limits for the data set size that Google is showing.
GSC LimitsFor example-GSC limits data to the top 1000 results- so for very large sites with over 1000 pages or keywords- one might have to pull data more frequently on smaller data sets to try to reduce the size to fit into 1000 results, but on these larger sites where there is a loss of available data- it causes incomplete data sets that contribute to some of the inaccurate results. However, for most smaller websites this factor usually doesn’t come into play. I don’t have all the answers as to why discrepancies or inaccuracies exist, but here are some insights that help improve some of the accuracy on data information you can be using. Overall GSC does give you information about keywords and organic landing pages- to provide strategic insights to help clients. But let’s look at some ways to better understand the data, and to get more accurate slicing of the data.
Why does GSC position data seem so inaccurate?It all depends on what GSC data you are looking at. There are many ways to slice and dice the data to see things more accurately in GSC. All results shown below use the same keyword example. By default if you do not use any filters the results shown for one keyword average together all the pages, all the countries, all devices and search types. So by default you are viewing a large aggregation of data sets. The keyword in Figure 1 received 37 clicks with average position of 3.3. We are only filtering on the keyword. This position information appears to be very inaccurate. Typically, my area of interest is to view US data, but is GSC really that inaccurate?
Figure 1- Default shows Worldwide all devicesIf I add a filter and compare devices- Figure 2 you see worldwide results as 37 clicks spread to 28 clicks for desktop and 9 for mobile- but receiving slightly better average position on desktop- but here if you proportionally average the desktop and mobile position results using the impressions- you get exactly 3.3 from the Figure 1 above, for all devices worldwide.
Figure 2- Shows Worldwide data set comparing desktop & mobileIf I then narrow the size of my database with a filter on just United States, Figure 3- I get fewer clicks over all, but you can see I also have an average position of 1 for both mobile and desktop in US. This compares with what I see in US. So, this filtering helped improve the accuracy of position data. Does this mean the views above are inaccurate? They do not seem to be inaccurate for the data sets they are including.
Figure 3- Narrow Data set to US, still comparing desktop and mobileNow if I back up and pick 2 countries to compare click and position results. Figure 4 shows US and France for all devices. But you can see I left out data for all other countries in this result- since here we have only 15 clicks and worldwide from Figure 1 we see 37 clicks worldwide.
Figure 4- Comparing Click & position results for US and France
Why do GSC landing page results show differently from GA results?This is one area that I have not fully figured out on why all the discrepancies exist. However, if I view GA as the “gold standard”, then my recent experience researching things for this article- shows GSC to be within +/-15% of GA landing page results. My research in the past showed these results to be closer together. I am uncertain how this gap has widened. So not knowing the factors- means other people may experience differing results. I believe even with this difference between GA and GSC, that GSC still provides valuable strategic insights for which keywords are bringing results for landing pages. In GA one needs to setup a segment “Sequences” view that filters on organic results where the source field is exactly equal to google- this will eliminate all other social, referrals and other things that might try to attach itself to the Organic search results. Then this becomes a fair comparison with GSC, along with defining specific dates for comparison.
Here are some discrepancies that can enter the data sets because of user actions-Possible Scenario for Discrepancy giving GSC more results than GA. A user searching for keyword 1 in one tab, opens a new browser tab enters keyword 2 and comes back to same website. This counts as 1 GA session if done within 30min recording only the 1st landing page in GA. But GSC is going to get 2 different landing pages and 2 different keywords. The GSC totals would be higher in this case. Possible Scenario for Discrepancy giving GA more results than GSC. It appears that ecommerce may cause multiple landing pages and sessions for 1 Search coming from GSC- see article talking about split sessions. When is a Session Not a Session? However I am still puzzling over my non-ecommerce situations where GA shows greater results than in GSC. Both scenarios can contribute to different GSC and GA results, which might account for some of +/-15%. However, there must be other scenarios that cause other differences.
Why does GSC only show partial keyword click results?If for the moment, we are comfortable using the GSC results even if they vary slightly from the GA results. A good portion of keyword click results are not included. For example, if I use the results to filter on a client page- as in the example in Figure 5, we see 133 clicks going to this landing page.
Figure 5-GSC click results for one pageThen selecting the Queries button, as shown in Figure 6, we see only 46 clicks for different keyword queries. This shows that only 35% are known and 65% are unknown queries for this page- meaning they are either “not set” or just “not provided” by Google.
Figure 6- Query results for 1 filtered pageThe estimated clicks column is how I make an estimate about a more likely resulting set of keywords for this page. By proportionally adding clicks back to equal the overall 133 clicks for the page. My assumption is that these same keywords that are shown to us have a high probability of providing a similar ratio for the remaining “not provided” clicks.
GSC Reliability Conclusions:
- Position data might be more meaningful to clients by filtering on just the segments of interest. Worldwide position data may not be as meaningful.
- GA and GSC landing page comparisons do not correlate at 100%. One might have to live with these inconsistencies- to get strategic value from the keyword results.
- GSC provides only around 35-50% of the keyword results for landing pages. It may be reasonable to proportionally back to these same keywords to fill in some of the missing information.