Audit – Update to the aggregate LinkRisk scoring method
For nearly a year now we have discussed internally the scoring method we use to derive the aggregate LinkRisk score for a profile within our Audit application. Today we are changing the way that is tackled by our scoring system.
If you are currently measured by your client base or employer on the risk score your Audit profiles show then you need to be aware of this change as it will have a significant effect on those scores.
The score we are talking about here is the one that shows in Audit when you run a profile as the overall risk metric for the site as a whole. It appears here: –
When we originally launched Audit the logic to calculate this number was: –
Take the average LinkRisk score for every active link and display that as the aggregate score.
With hindsight that meant this score had a few issues for me and so I started a discussion internally to push to have it changed. The main issue was that if we took the average of every active link that meant that we could see the overall score being pulled one way or another by the presence of a sitewide link from a trusted or a bad domain. You could find yourself in a situation where having a link on every page of your clients head office would mean that the overall average for the domain was pulled down into low risk, masking the actual issues that may exist elsewhere.
Still with me? 🙂
OK so what is the solution?
Today we change the logic we use to make the score more accurately represent the true risk of a profile. To do that we now use the following logic to derive the aggregate LinkRisk score: –
Take the mean of the links pointing to a domain and then calculate the mean of those averages.
This basically means that we take the average of all the links from each domain and then each domain only passes one vote of risk to the main calculation. This means that we no longer have the situation where the presence of a sitewide can overtly influence the overall risk profile of an audit.
A word of caution
We still strongly advise our customers to avoid using the aggregate LinkRisk score as a metric to track and be measured on as an indicator of progress. There are many factors that influence that score: –
- Disavow activity on the domain
- Link removals and cleanup
- Gaining new links
- LinkRisk algorithm updates
- Crawl memory from external data providers
An example of how relying on that number as your main benchmarking metric could become problematic might be as follows. You manage to get a profile down by 30% through cleanup and disavow work and then we release an algorithm update that targets a specific tactic and that means that the profile then reverts back by 20%. This isnt an indication that you’re not doing your job, its an indication that we got better at ours and together we have some more work to do to make things even safer.
As with all of our scoring updates these changes will be in place when new profiles are run or an existing profile is rescanned.