EDIT: I got every part of this analysis right, up until the very end when I had simply to average some figures out. Please see here for the updated version. Apologies.
I’m on holiday today, so I thought I would spend the afternoon drinking coffee, listening to Pandora (Nick Drake Radio – perfect), stroking the cat, and delving into the metrics I use to assemble the PowerPR index to see which is the most influential. Do I know how to party or what?
This is something I discussed a while back with Flemming Madsen of Onalytica. We had an interesting conversation about what exactly constitues a PR blog. Does it have to talk about PR exclusively? Should it just be an individual’s blog, or should it include blogs run by companies? And, most intriguingly, which metric holds the greatest sway over a set of combined rankings?
In the PowerPR index I use a variety of metrics. Some could conflict with each other, while there may indeed be overlap between them (the subject of another future post). My attitude is, if it’s available, throw it in the mix. This is in the absence of being able to find any appreciable correlations between them which is what got me started on all this in the first place.
So, to the metrics. I use Technorati Authority, Rank, Inlinks, Yahoo Inlinks, Google Hits, Google Blog Hits, Google Blog Hits over the past month, Blogpulse Hits, and Blogpulse Hits over the past 6 months. Which of these is the most influential? That is, if I rank the PowerPR index according to just, say, Technorati Authority, how close is it to the ranking obtained when looking at the metrics combined? If it’s identical then it implies Technorati Authority holds total sway, and my PowerPR index is knackered.
Fortunately, that’s not the case. A list ranked by Technorati Authority alone is similar, but not identical, to the full list. Here it is:
As you can see, there are broad similarities, but significant differences too. First Person PR, for example, would shoot up 20 places if I just used Technorati Authority (btw, all the differences are absolute differences because I’m interested in the change rather than the direction). Unfortunately for Kari Hanson, I don’t.
The ‘average difference’ (you can tell I’m not a statistican, can’t you?) for Technorati Authority is
9.4. This means that, if I ranked purely according to Technorati Authority, on average a blog would change its ranking by nearly ten places. This is quite a difference and implies to me that Technorati Authority alone isn’t well aligned to the rankings when looking at all metrics combined.
Post-edit: This is where it went a bit wrong. I strongly recommend you continue from the more recent post where I fixed my figures – click here to see it.
So let’s look at the other metrics. I’m not going to reproduce a table of the whole lot, but the ‘average differences’ are: Technorati Authority=9.4 Technorati Rank=49.21 Technorati Inboundlinks=12.44 Yahoo Inlinks=1.51 Google Hits=11.12 Google Blog Hits=9.07 Google Blog Hits over a month=14.85 Blogpulse Hits=15.82 Blogpulse Hits over 180 days=14.54 It looks to me like Yahoo Inlinks is the single most influential metric. If I rank according to Yahoo Inlinks, I will get a very similar table to when I rank according to all the combined metrics. Technorati Rank is way out there. A ranking by Technorati Rank alone would almost completely change the table. This implies to me it’s measuring something very different from the other metrics. On this basis do I keep it – because it’s measuring something different from the others and is therefore valuable – or ditch it because it’s screwing the table up? I think the key to this is the ubiquity of Technorati Rank. Every blog has one, so by including it at least I’m being fair, which other metrics such as Alexa might not do because they’re not universal and have a tech bias. It’s peculiar however that the Technorati metrics especially display such variance. Why should the ranking be so different from Authority and Inboundlinks? Quite possibly the subject of another post…
I should, of course, qualify everything I’ve said here by reiterating that I’m not a statistican. I just quite like data mining and looking for patterns (I’m a pattern-matcher by nature, as I’ve found out in tests that measures people’s heads in the past). Also, whereas I’ve tried *really* hard to make sure all my data is good, this kind of analysis is pretty complicated and I do hope I’ve managed to copy and paste everything in the right place. Let me know if you violently disagree with my findings!
POST EDIT: I obviously didn’t try hard enough, and I do wish people had violently disagreed, because I was wrong…