Rank Tracking Experiment: Negative Queries
Posted by IrishWonder
on April 25, 2018
Do negative queries have influence on rankings of a site? I decided to run a test and check if rank tracking for a phrase like [generic keyword] -[brand] would influence the rankings of said brand in the [generic keyword] SERPs. If this seems like I’m jumping ahead of myself a bit, I’ll explain:
- depending on how rank tracking is implemented, it may or may not be perceived as actual human users searching for a keyword;
- it has been speculated before that rank tracking tools can influence a keyword’s reported search volumes;
- in the past I have run campaigns with the purpose of changing the search suggestions by means of, among other things, inflating the search volume of a desired suggestion with automated tools posing as human user agents – no, Google can’t tell the difference;
- I am not the only one who has done this – infact back in the day at least one SEO has been told by Matt Cutts to cut down the MTurk activities;
- for the sake of avoiding unnecessary exposure, I am not revealing the query or the brand involved in this experiment – I will just refer to them as and [brand].
Prerequisites: I picked a mid-tail, commercial intent keyword where brand X has been historically ranking on the first page quite stably. I set up rank tracking for two queries:
Here is what I observed:
The top 100 for (white spaces stand for proxy bans):
The top 10 for – notice the total variance:
[brand] site rankings for during the experiment – the lowest point is moused over for clarity:
The top 100 for -[brand] – notice the greater variance in individual URLs’ rankings:
The top 10 for -[brand] – notice how the total variance differs from that of the top 10 for :
After a while, the negative query tracking was stopped – below are the brand site’s rankings for over the period when no negative queries were tracked:
What to make out of it all: I cannot 100% ascertain that running negative queries affects a site’s rankings. This is a sample of one query in one SERP over a limited period of time so probably statistically useless. The brand site’s drop to No. 5 at one point could be influenced by my experiment, or it could be just a random fluctuation. The more stable graph of the brand site’s rankings outside of the experiment can prove either point. However, what we cannot ignore is a much greater variance of the URLs in the SERPs for negative queries than for regular queries – it’s almost as if Google, if told to serve up the results for the query but not the one result it has always thought belongs there, does not really know what to serve up.
Feel free to share your thoughts below in the comments.