What I did was take three important keyword phrases for our local clients, get the top 20 local results, and compare those on these factors:
For the Google local portion:
- Reviews
- Number of incidences of parts of keyword in local business listing title
- Photos/videos
- User content
- Webpages
For the destination URL:
- Inbound links
- Pagerank
- Number of pages in site
- Number of incidences of parts of keyword in homepage title tag
With all that data and my very naive statistical abilities (am I the only guy who wishes he had a statistics professor chained to a dingy, cobweb-ridden cubicle in the corner?) I created scatter graphs and found R2 values.
Unfortunately, the strongest correlation is still considered "weak" in statistical parlance. I suspect someone from Google will read this and laugh at my non-PhD-ish bafflement. Nonetheless, I know where to go next- so there will be a part two.
Strongest correlation: More reviews equals better ranking
The correlation of two important SEO factors... also weak:
Other factors and correlative values (all weak, but strongest first):
- Number of photos and videos (listed in local details): R2=0.2053
- Number of webpages (listed in local details): R2=0.1862
- Number of user content (listed in local details): R2=0.1788
- PageRank: R2=0.154
- Keyword element instances in local listing name: R2=0.079
- Keyword element instances in homepage title tag: R2=0.0011
Next up, I'm going to examine:
- More keywords for # of reviews, larger dataset to get more certain R2 value for this
- Keyword prominence for homepage title and LBL name
- Keyword integrity (whole keyword in order) in homepage title and LBL name
Cool stuff Brian. I recommend you throw in “Web references” for biz as web citations appear to be playing a part in the Google Local Algo
What do you mean by # of webpages listed? You’re talking about GoogleLocal right?
Web pages means when you’re looking at the business listing in maps.google.com (where you can write reviews), it’s the rightmost factor with a number in parentheses, to the right of “photos & videos” and “user content”. Andrew, I’m assuming you mean the same thing?
Yeah, try it with what Andrew said 🙂
Nice post by the way.
ummm, nevermind I see you already did, R2=0.1862
was too caught up looking at the pretty graphs
Oh, cool. thanks!
Very nice study Brian. I took part in Mike Blumenthal’s and David Mihm’s research last year.
Location seemed to be incredibly important for local businesses with less signals (citations/web pages, reviews/user content, pics, etc.).
Citations seemed to be very important with sites w/ more signals.
What kinds of businesses/industries were you reviewing?
Btw; from just observations …and specifically looking at the spamologists that have taken over the locksmith industry….it sure looks like reviews have lots of importance.
Earl, this was specifically for our local tourist destination industry. Only three distinct keywords, but popular ones. Next I’m going to go into more keywords on fewer ranking factors. Thanks! 🙂
I find it curious how “Keyword element instances in local listing name” is so low, R2=0.079
I’ve seen cases where the only change that was made to a listing was adding a keyphrase to the business listing name, as a tagline where proper name gives no clues whatsoever to the businesses activities, allow a listing to shoot up into the 10 Pack within days. Granted, competition with other key factors, mainly reviews, was very very low.
Hi Brian,
One big factor not to overlook is that of the trust that Google has in the information they have about a business. It’s not just the number of the web pages they find about you, but the consistency of the information that’s on those pages and the authority of those pages in Google’s eyes. I don’t know how easy or hard it is to study that statistically, but go for it! Thanks, Mary
This is great and timely stuff.
As to consistency, as Mary referenced….from a different perspective…RIGHT NOW…and over the last couple of weeks, Google Maps is struggling with merged records…with data from different sources being merged creating false information. On one case reported by Blumenthal both he and Google personnel manually had to take a hard look at a record to ascertain if the bad information was either merged or hijacked.
I believe Google Maps methodology and hence its rankings are still in their infancy as Google works through bugs. I suspect that ranking in maps will be a moving target with different influences taking precedence as we move forward.
More reasons for more current studies….like that of Brian’s.
For those who recall….Google’s organic rankings and visibility of local businesses absolutely stunk until they made a dramatic algo change implemented in early 2005 and reported in a review of patents by Bill Slawski in latter 2005.
I’d like to see them work through crummy and falsified reviews in the interim. That is incredibly manipulative…as is being shown by the spammers.
Ya, I didn’t want to mention it, but some listings could be hijacked AMAZINGLY EASILY- I’m certain we have local businesses who love the listing they have (whether they look at their analytics or not to see how it benefits them), but since they haven’t claimed it, it’s vulnerable.
Reviews as the best :signaller:?
Hmm, I’ve held that opinion almost permanently.
Yet, looking this morning, we put up a new resort
in NY and the search ‘name-of-town resort ny’, and permutations, showed Local Listing yet with only 5 reviews (it’s been open for only a few weeks). I was very happy, but very surprised.
Nice to see this pop up on Sphinn.
As EarlPearl mentioned above, if you’re doing a second round of testing, be sure to include distance from centroid in your factors. While it appears that this has less effect in populous areas and in well-documented industries, the effect still definitely appears to be there in more rural areas and less crowded industries.
I’ll be interested in Part 2, Brian, and will give this a sphinn.
Very cool post. I too am a statistical handicap but to know that there are factors that influence ranking more than others is good. In my small market it doesn’t seem like any listings have been hi-jacked but I’m just waiting for it to happen, the majority of listings in our market haven’t been claimed psshhh… When will they learn…
what i see is the pages to inbound links ratio pages 150-250 pages = 750-1000 inbound links is that right.
LeeLow, that’s probably only applicable for this limited group of searches- check out Mike Blumenthal’s study of more data until I get a larger sample size for you: http://blumenthals.com/blog/2008/08/04/ranking-factors-in-google-maps-cracking-the-code-smx-local/
Brian,
Excellent work, I was getting really annoyed with the anecdotal evidence from the “experts” and what they “believe” impacts the ranking. Your analysis here correlates with my own anaylsis. No clear a clear factor is not presenting itself. I intend to look deeper at factors such as whether the listing has been claimed by the business and the PR (authority) of the sites that reference the business as well as the age of the business llisting (working on how to measure this one). Anyway, looking forward to your follow-up report, keep up thegreat work..
Thanks Rich. I looked at more data points today, and still, almost any single factor appears to have a weak correlation.
However, I think this makes sense when an algorithm is a weighted multi-factor ranking… maybe it’s ok that they’re all weak alone- and it does help us prioritize what to work on first.
I’m going to keep looking at the data. Found a few interesting things about keyword in LBL name and category that I’ll publish soon.
To make it all the more confusing, I’ve seen the raking change in successive searches. Not sure if it’s the personalized search heuristic or a touch of randomness thrown iinto the mix.