Does anyone have a 500,000 daily pageview site?

1 Trackbacks to “Does anyone have a 500,000 daily pageview site?”

1. Trackback from Dan Grossman » Where can I buy unique articles? on October 25th, 2007 at 1:43 am:
   

   […] volunteers for my testing request […]

14 Responses to “Does anyone have a 500,000 daily pageview site?”

1. Ryan Williams October 22nd, 2007

   Hi Dan, Not really related to this post, but I was just wondering: does W3Counter have anything these days to prevent server load on the W3C end, even if unlikely, from slowing down the site being tracked? I remember this being a huge problem in W3C’s early days, and that has made me really cautious about using externally-hosted statistic software. It pretty much made my sites look like they suffering from downtime, despite the host server being fine. I do recall seeing a comment here ages ago that described the possibility of hosting the JavaScript and tracker image ourselves, which seems like a good solution. Could some proper instructions appear for doing this at some point? Also, I don’t know if the “What makes it fast?” part of the WordPress Official Stats article is relevant: http://andy.wordpress.com/2007/05/05/automattic-stats-for-self-hosted-wordpress/

2. Ryan Williams October 22nd, 2007

   BTW, is the referrer report *meant* to count referrals from your own pages? Pretty much my entire referrer report is currently filled up with various pages of my WordPress installation, drowning out those who’re getting referred by external sites. Seems like really weird behaviour, anyway.

3. Dan Grossman October 22nd, 2007

   Ryan, you’re using CSS to move the counter image off the page instead of paying for the invisible version, while sending half a dozen feature requests a day? You want me to build this service for you, do it without slowing down your page, and you want it without paying and without even giving the link back to W3Counter, which is all I ask? Get outta here.

4. MaxS October 22nd, 2007

   Cheers Dan! The comments gave me a nice laugh. I love the W3Counter widgets, by the way. Way to innovate.

5. Robert Norton October 22nd, 2007

   Dan, I’ve got a feature request, a bit OT though. How about a feature in your admin portal that lets you block/remove active sites for a particular user ;)….

6. ses5909 October 23rd, 2007

   Oh i was coming to omment on the Linux posts and they are gonnnnnnnnnne! I wish I had that busy of a site. Sorry, nothing useful to add here, as usual

7. Ryan Williams October 23rd, 2007

   Sorry Dan, but I’ve no intention of actually using the service at this point (Mint is already tracking that site) and I was just trying to look through it properly without messing up the design of my page, which it does by default; if I were to continue using W3Counter beyond a couple of days I’d have immediately paid for the invisible version. I’ve removed it now as the things I’ve observed prohibit me from really rolling it out across my serious sites. I apologise if you saw this as cheeky, but in all honesty you weren’t going to get anyone coming via that image for the day or so I was planning to keep it there anyway. If you don’t see my suggestions as worth following up on then by all means, don’t listen to them. It’s just that functionally this application seems to have evolved little from its very first iterations back in 2006 (including very obvious issues, some of which I mentioned back then too), and I can only put that down to you not getting enough feedback on areas where’s it’s lacking. Is this affecting sales? You tell me. Sorry for the offence!

8. Dan Grossman October 23rd, 2007

   Well, thanks for the feedback. Don’t know what you mean about messing up your design though. Haven’t seen it mess up a site’s design yet?

9. Ryan Williams October 23rd, 2007

   Perhaps ‘mess up’ is too strong a phrase for the reality, but it does cause the normally-unseen huge block of green that’s the site background behind the main holding DIV to show up at the bottom, which as someone with mild OCD bothers me quite a bit. :~ I realise I should have probably spent a little longer remembering how I put together the template a year ago and working the counter in properly, but I honestly didn’t consider it as worth the time or think it’d be noticed as I was only trialling it before making a decision to use it on other sites (I’d continue to use Mint on that one seeing as I’ve already got an eternity of logs built up). I’ll take the slap on the wrist and concede that I shouldn’t have done it. I didn’t mean to piss you off by doing it, and perhaps I can make it up in the future by subscribing to W3Counter some day. After all, I’ve been checking in on it every so often for a very long time to see if the key niggles have been addressed (local referrers being counted is a major one I’m sure I reported in 2006). I do want to subscribe — honest! I just can’t live with certain things.

10. Dan Grossman October 23rd, 2007

    But your site is the referrer of those visits! It’s a consistency problem. Half the reports (top referrers, referring domains, new referrers) would ignore those visits from your site, while half (online now, online recent, recent referrers) show them and show your site as referring the visit? It can’t lie and say only 10 people are online when there are really 15, 5 of them starting their visit from a browser window left open at your site for an hour or two. It also means the percentages for all referrers/domains will no longer add up to 100, since some percent was referred by items being hidden from the list. Is this actually preferable?

11. Ryan Williams October 23rd, 2007

    Why not just stop counting them as referrals? You could approach this by labelling visitors who’ve gotten there from another page on the same domain with something like ‘Local Page’ where the referrer is shown now. This means those visitors wouldn’t appear at all in the referrer-specific reports if they weren’t referred externally, and for the ‘Who’s Online?’ pages and such they’d still appear but just with ‘Local Page’ or something as the referrer (you can probably think of a better term than me). I genuinely believe most people would prefer they’re not counted as referrers at all rather than local referrals filling up the referrer reports. When I look at a referrer report my intention is to see which sites referred people to me; I have absolutely no interest if they got there from my own site’s pages, and I can’t be the only one thinking this way. The way it is now *almost* makes it no better than just reading a generic long report of every visitor rather than a report tailored specifically towards drilling down what sites have referred visitors.

12. Dan Grossman October 23rd, 2007

    Try it out then. http://beta.w3counter.com/stats now does all you’ve asked. Or if you’ve already removed the site, you can compare these: http://beta.w3counter.com/stats/referrers/1/top/20/0 http://www.w3counter.com/stats/referrers/1/top/20/0 All reports that deal with referrers (top referrers, referring domains, new referrers, recent referrers, online now, online today, visitor detail, visitor by label detail, top referring domains widget, pwidget referrers tab, dashboard sources list) were updated to either filter out or rename visits where the leftmost part of the referring URL matches the URL of the site entered when it was added to W3Counter. But man does this cost… the queries which group by referrer now can’t filter on the unique visit index, so a larger index or a row scan has to be used, which makes the query slower. Then a second query has to be performed to add the self-referred visits back in to the “direct visit / bookmark” row’s counts so that the totals add up to 100% again… that query also can’t use one of the smaller indexes, so I add about 4/10ths of a second to the report for that… with ~30-40 concurrent users logged in and viewing reports on average, that’s a lot of extra memory and cpu cycles that can’t be used on logging new data. Hrm.

13. Ryan Williams October 23rd, 2007

    I see the performance concerns (definitely significant when you consider the bigger picture and expected growth; no suggestions I can think of there beyond pricey clustering), but I definitely think it makes the whole concept of a referrer feature more useful to webmasters. I mean at the end of the day, anyone wanting referrer statistics is primarily going to be interested in seeing which external sites referred visitors to them. I’d be very surprised if more than a handful of people really want to know what internal pages people have arrived from, and if they do want to see a visitor’s path through the site they can still use the ‘Online Today’ and ‘Online Now’ reports to get a full view. I’m pretty sure this is the approach Mint and Reinvigorate use, as well as other statistics programs I can recall using with referrer data; stripping out local referrals seems to be pretty much the de facto standard if a separate referrer report is offered. They also strip out ‘Direct Visit’ links so they’re not shown in the referrer report either; I notice that on your beta version these are still shown, which makes no sense if there’s no referrer at all. I realise this might be a bitch if your programming can’t really deal with it as optimally, but I’m convinced what I’m saying is right, logical, and would improve the usability. Seeing where people are coming from is something a lot of webmasters take keen interest in, so making that experience as easy as possible is always good. I know I’m verbose and I apologise, but the crux of my argument here is that the term ‘referrer’ is most usefully interpreted as ‘another site’s page that sent a guy to mine’, and this is what people primarily expect to find; anything else such as direct visits and internal referrals are just superfluous. All I can urge you to do is continue looking into this direction, anyway; I think the changes made in the beta are already a huge improvement. You never know, what I’m saying might be exactly what turns some people off who’re looking for referrer data presented how they’re used to. :> By the way, that pairing of usernames with statistics is very sneaky and very cool (I didn’t test W3C long enough for it to pick anyone up). Great work.