Ask HN: How accurate are Alexa/Compete/Quantcast for your website(s)?
16 comments
I've got some sites that are 'quantified' with Quantcast (have the tracking code) and the numbers agree within a few percent with other analytics: that's for sites that get anywhere between 50 visits a month to 500k.
If you're not quantified, Quantcast is a joke and the other sites are always a joke, no matter what. There was a long time when Alexa thought the top 20 sites in the world were all in Korea because, I guess, everybody in Korea had the Alexa toolbar. One day they changed their formula to prevent this because people were pointing out that their stats make no sense.
Compete, in particular, has a free service that's designed to annoy you into paying for the service, but everything I see on their free service and on their blog convinces me that the paid service is worthless... For instance, they had nothing to blog about one day, so they blogged that "Girls Gone Wild" gets 30% of its traffic from women... Everybody knows that systems like that use 50-50 gender statistics as a bayesian prior, so they always claim that sites have more balanced audiences than they really do.
If you're not quantified, Quantcast is a joke and the other sites are always a joke, no matter what. There was a long time when Alexa thought the top 20 sites in the world were all in Korea because, I guess, everybody in Korea had the Alexa toolbar. One day they changed their formula to prevent this because people were pointing out that their stats make no sense.
Compete, in particular, has a free service that's designed to annoy you into paying for the service, but everything I see on their free service and on their blog convinces me that the paid service is worthless... For instance, they had nothing to blog about one day, so they blogged that "Girls Gone Wild" gets 30% of its traffic from women... Everybody knows that systems like that use 50-50 gender statistics as a bayesian prior, so they always claim that sites have more balanced audiences than they really do.
Wikispaces.com does 9M uniques and 75M pageviews a month per Google Analytics. Quantcast says 1.4M uniques, Compete says 900K. Neither trend in the right direction.
Alexa is comically low on pageviews and maybe 6x low on reach when we compare those stats to other sites who make their numbers public. However, Alexa tends to head in the right direction.
I looked at Comscore data on us a few years ago and it was less wrong than Quantcast and Compete, but still well off the mark.
Alexa is comically low on pageviews and maybe 6x low on reach when we compare those stats to other sites who make their numbers public. However, Alexa tends to head in the right direction.
I looked at Comscore data on us a few years ago and it was less wrong than Quantcast and Compete, but still well off the mark.
My experience is that they're all easily gamed if you want to.
As an example, I once spent a year browsing with the Alexa toolbar, and it boosted every one of my sites up into the top 100k just from my own traffic. After ditching the toolbar, they all dropped back down.
As to accuracy, Compete isn't even on the right order of magnitude for any of my sites. It's way over for one, way under for another. Just no correlation to reality.
As an example, I once spent a year browsing with the Alexa toolbar, and it boosted every one of my sites up into the top 100k just from my own traffic. After ditching the toolbar, they all dropped back down.
As to accuracy, Compete isn't even on the right order of magnitude for any of my sites. It's way over for one, way under for another. Just no correlation to reality.
CodeAnthem.com has been up a few months, so only full month avail was April on Compete:
Google Analytics (unique visits): 77K
Compete: 20K
Quantcast: No Data
I really have no idea how to find comparable stats on Alexa (since it uses percentages) but you can check it to see.
Compete's referral data and search analytics were also way off. I might think this is due to the site being pretty new ... but I'd rather they said nothing (like Quantcast) than be way off.
Based on how wrong they are for my site, I would not put much stock in them for other people's...
Google Analytics (unique visits): 77K
Compete: 20K
Quantcast: No Data
I really have no idea how to find comparable stats on Alexa (since it uses percentages) but you can check it to see.
Compete's referral data and search analytics were also way off. I might think this is due to the site being pretty new ... but I'd rather they said nothing (like Quantcast) than be way off.
Based on how wrong they are for my site, I would not put much stock in them for other people's...
If you're "quantified" on Quantcast, it should be pretty spot on to Google Analytics since it's direct measurement. If you're not, then it's definitely going to be off like Compete.
I use Compete stats as relative measurements against other websites (assuming everyone's off by a similar ratio). I use Quantcast stats to look at the demo info.
I use Compete stats as relative measurements against other websites (assuming everyone's off by a similar ratio). I use Quantcast stats to look at the demo info.
Somewhat random fact: comment solutions like IntenseDebate and Disqus incorporate Quantcast into their js, so you are indirectly getting counted by such services if you use IntenseDebate, etc. I searched for an out-clause, but found none.
Could you please expand upon the implications of this? It is unclear to me whether this is positive or negative aspect for the different entities involved: the commenting apps, Quantcast and the website they are embedded on.
I checked my quantcast stats a few weeks back after taking off their js months ago. I noticed that they still had a good picture of views/etc and wondered why. It appears that they were getting info from Intense Debate.
I originally took the Quantcast code off my site b/c it slowed things down and despite having fairly accurate counts, it had skewed demographics. I didn't want to encourage them, so I took off the script. However, my use of IntenseDebate meant that I was still using it.
I don't think IntenseDebate is doing anything wrong, but I would like an opt-out.
I originally took the Quantcast code off my site b/c it slowed things down and despite having fairly accurate counts, it had skewed demographics. I didn't want to encourage them, so I took off the script. However, my use of IntenseDebate meant that I was still using it.
I don't think IntenseDebate is doing anything wrong, but I would like an opt-out.
I would urge people to include Comscore as well. None of these sites track accurately. Here's a site I used to work on.
Site: Gameriot.com Month: May '10 (May 1 - May 31) Google Analytics: 307,000 "Visitors/Month" (Unique Visits) Alexa: 25,856 Traffic Rank (has been as high as 2,000) Compete: 72k "Unique Visitors" Quantcast: 201.5k 'Worldwide Visitors' Comscore: 144k "US Visitors" (we have to pay more than $30k/yr for global access)
Site: Gameriot.com Month: May '10 (May 1 - May 31) Google Analytics: 307,000 "Visitors/Month" (Unique Visits) Alexa: 25,856 Traffic Rank (has been as high as 2,000) Compete: 72k "Unique Visitors" Quantcast: 201.5k 'Worldwide Visitors' Comscore: 144k "US Visitors" (we have to pay more than $30k/yr for global access)
Site that's been around for years:
Google Analytics: ~1,000,000 visits/month
Compete: 110,000 visits/month
Alexa: 9,548 traffic rank?
Quantcast: 5,000 visits/month
They all seem to be very low and very all over the road.
Google Analytics: ~1,000,000 visits/month
Compete: 110,000 visits/month
Alexa: 9,548 traffic rank?
Quantcast: 5,000 visits/month
They all seem to be very low and very all over the road.
A theme I've noticed is that techie sites have their traffic grossly underestimated. If the figures are being estimated through traffic going through a handful of points, this makes sense to me as the visitors to, say, a Python documentation site are likely to be less diverse than those to a Flash games site or Facebook.
http://onlineslangdictionary.com/
-----Uniques (March)
Google Analytics - 470k uniques
Compete - 124k
Quantcast - 156k (not "quantified")
-----Rank (at present)
Alexa - 19,730 in the US
Compete - 16,735 (US?)
Quantcast - 12,957 in the US
-----Uniques (March)
Google Analytics - 470k uniques
Compete - 124k
Quantcast - 156k (not "quantified")
-----Rank (at present)
Alexa - 19,730 in the US
Compete - 16,735 (US?)
Quantcast - 12,957 in the US
As a corollary to this, how would people rate Google Analytics versus their own traffic logs?
Here are my small niche sites:
--------------------
Site #1 (classic gaming related)
April 2010 - Unique Visitors:
My server logs: 16000
Compete: 700
Quantcast: -
Alexa: -
--------------------
Site #2 (music related and random things)
April 2010 - Unique Visitors:
My server logs: 5000
Compete: 2326
Quantcast: -
Alexa: -
--------------------
Site #1 (classic gaming related)
April 2010 - Unique Visitors:
My server logs: 16000
Compete: 700
Quantcast: -
Alexa: -
--------------------
Site #2 (music related and random things)
April 2010 - Unique Visitors:
My server logs: 5000
Compete: 2326
Quantcast: -
Alexa: -
how do you find the uniques through server logs? I know software like webalizer and awstats overexxagerate quite a bit if left alone (count too many bots, extra uniques where they shouldn't be, etc).
Compete.com is US only traffic so it is pretty spot on for your second site. However it's weird that your first site with 500+ uniques/day avg. doesn't have a quantcast estimate.
Compete.com is US only traffic so it is pretty spot on for your second site. However it's weird that your first site with 500+ uniques/day avg. doesn't have a quantcast estimate.
One thing to keep in mind is that cookies are not people. People clear their cookies, so any reliance on that as a measure of unique visitors has the potential to vary widely.
In my experience, http://www.google.com/adplanner has been far more accurate than Alexa, Compete, and Quantcast.
Compete measures iframes and js includes, meaning any company with a moderately successful non-flash widget is drastically overestimated.
Off by an order of magnitude for many of my sites.
This post was inspired by a Reddit admin's comment showing that Quantcast was far off: http://www.reddit.com/r/reddit.com/comments/cahpw/digg_is_dead/c0r9904
Please post numbers for months that are fully shown at all those sites already, so not from this month. You don't have to post your domains for privacy reasons, but if you would, please do.