Statistics, Anyone?

Over the past couple months, and particularly over the last two days, it has become interesting to me what kinds of visitors I am getting to the site and where they're coming from. There are a few stats packages, and I've been working out their benefits and flaws.

I was able to jump on the Google Analytics train before they closed their doors. I installed the required script in the necessary places and waited for stats to roll in. And I waited. And I waited.

Eventually they had enough data to play with (do I not even register in the Google Analytics world as anything but a blip?) and they started showing me some statistics. But. How do you use this thing? Even being a former Urchin 5.0 user, I was still somewhat confused by the interface. Suffice to say, I never really figured it out, and I started to suspect that using Analytics was affecting my AdSense adversely (though I can offer no evidence that this is anything but paranoia), so I simply removed the tags.

Isn't there something that you can just push a button and get the stats you need? And exactly what stats do you need?I also have a copy of Mint that I use pretty regularly.

There are things I like about Mint. Peppers are great. I can add the functionality I want, provided someone writes the pepper for it. The default peppers aren't all that informative by themselves, but when combined with outgoing clicks and referrer data, it becomes very useful.

There are two problems I have with Mint, though. First - and this was something that I liked when I first started using it - I hate the UI.

No, really. It's fundamentally a pretty and elegant design. But I can't stand looking at it any more. It's a big long page of data and it's all green. Green! Ok, so that's just an aesthetic issue, but shouldn't the important data be right there in front of me? Yeah, sure I can drag things around to put important stuff near the top, but shouldn't this type of software know what stats I'm going to be interested in and put them in front? Ok, so maybe that's just me - but my second issue isn't.

There is no historical data in Mint. Or very little. I haven't figured out a way to see trends over a longer period than the "right now" that Mint shows me.

What was the most frequently searched term yesterday? Last week? Last month? Year? Who knows?!

It's possible that Mint is storing all of this information somewhere that I can't see, and that I can access it via a pepper, but I haven't found the pepper that releases that functionality yet.

There is at least one WordPress plugin that tracks statistics, too.

StatTraq is a WordPress plugin that tracks statistics. It seems to create a new record in the database for each hit (which is to be expected) but never cleans up any of it. So I can get historical data as granular as a single hit, at the expense of a pretty darn large database. I haven't looked very deeply, but although I expected that there would be some data cleanup, I could not find the SQL "DELETE" statement anywhere in the code.

StatTraq also doesn't seem to record hits for pages outside of WordPress. Once again, I'm not sure if this is specifically true, but I don't see a way that the code that records hits can be executed from outside WordPress.

A nice thing about StatTraq is that it is integrated with WordPress. It stores actual post IDs, not just the URLs, so you know when a specific post is being read whether though a friendly permalink or the crufty "?p=2482" URLs that sometimes sneak by.

StatTraq also tracks a good footprint of tracking data, which you don't get from simple 3rd-party counter services that display a simple graphic number at the bottom of the page.

I don't really trust 3rd parties with my stats. Even in the case of Google, I was worried that they were using my stats to influence the ads placed on my site. It could be paranoia, but if the stats are hosted locally, then I don't have to worry about any of that.

Of course, there are log-reading scripts that will accomplish similar tasks by running on your server. Webalizer is an example of such a script.

What makes scripts like Webalizer different from Google Analytics or Mint is that they don't collect data on their own, but use the existing server logs. This has benefits and drawbacks.

One benefit is that the web server usually keeps the most accurate account of what files it served. Using that log, you should be able to create a perfect picture of exactly what requests were made. This is particularly so when comparing Webalizer to Google Analytics or Mint, since both of those products rely on Javascript. If a visitor has javascript disabled, or a version of javascript that can't execute the logging code (like in a mobile phone browser), then no hit is recorded!

Another benefit is that because the server is likely already storing these logs, it doesn't take an additional chunk of space to store hits in a database. Usually programs like Webalizer generate their reports at set intervals as static HTML pages of results. And that brings us to the down-side.

Because Webalizer isn't typically run in a "live" mode, you don't see real-time statistics. Instead, you get periodic snapshots of particular ranges of time. Perhaps that's enough of a view of statistics for you, but I prefer to see trend information as graphs over time, not just an aggregate snapshot of data during a range of time.

Also, because it's not using anything in addition to the standard server logs to hold data, the results typically lack any information that you can get by using the javascript used by other programs. For example, screen resolution of the browser is not something typically stored in the server logs. If you're looking to your stats to determine the best size for a redesign of your site, a program like Webalizer isn't going to provide that information.

Another possibly troubling aspect of using only server logs is that there really isn't a great way to reliably track sessions (what Webalizer calls "visits"). Knowing the number of unique visitors versus the number of total hits can tell you if people are reading more than one page from your site when they visit. This is important if you care about "stickiness".

Plus, aren't those stats kind of ugly? I guess they're functional.

AWStats is another stats package in the group of log analyzers. Once again, it's relying on the server logs for periodic publication of reports. It does look a whole lot better, though.

There are a couple of other projects worth metion in statistics tracking.

BlogBeat is a relative newcomer to the statistics arena. The project is still in development, although it seems complete.

Blogbeat uses the javascript method (like Google Analytics) to record statistics information. It is also a 3rd-party service, so if you're uncomfortable about your stats are being fed to some other site, this one also isn't for you.

Blogbeat has a very simplified interface for providing just the stats that it thinks you will want. I like the idea of reducing the clutter of overwhelming and often incomprehensible information provided by stats packages. I'm all for it. Still, Blogbeat seems a tad light on per-page content.

Something I really like about Blogbeat is the integration of FeedBurner statistics. Feedburner has an API that lets you obtain minimal information about feed readers, and Blogbeat integrates that into its graphs. That's pretty cool.

Something that bugs the hell out of me is how Blogbeat uses links to itself for every logged page. For example, imagine that it's telling you that 300 people read the page titled "A Nose to Honk By". You don't remember what that post is about at all You want to read that post to see what they're talking about. So you click the link. Guess where it doesn't go. In fact, there is no link that goes to that page of your site. There is a URL displayed, but it's not linked. Perhaps this is just a small oversight.

Blogbeat has reasonable granularity. It can give you statistics for gradually longer date ranges - Today, Week, Month, Year, All. So you can find out what the most popular post is for this month or last month, but not the month before that, unless you also toss in all the popular posts for the last year. I don't see this as a problem, really, and if you're looking for more detail than that, then you probably need a higher-end service.

Also - Everything I've mentioned so far has been free to use and/or install. Blogbeat is free for the first 30 days, then it's anywhere from $6 to $79 per month, depending on how many hits your site gets.

Another option for stats is Measure Map, which seems unavailable to the public yet. Most people who have received invitations to use it have been saying good things about it. I signed up for an invite months ago and haven't heard anything, so I suppose I've been snubbed by automation.

From what reviews I've read and screenshots I've seen, the nicest thing about MeasureMap is that it's pretty. There are fewer endless dull charts of itty-bitty stats crap here, and things seem visualized a bit better than they are with other packages.

But I haven't personally reviewed the software, so I'll reserve any opinion until that happens.

So what's all this talk of statistics about, anyway?

I'm trying to come up with some dream features and the technical requirements to support them.

When you're looking at stats, why are you looking at them? What are you hoping to see?

I've had some people already tell me what they use stats for. Here are a few things that I've learned:

  • People want to see who is linking to them and which of those links are being followed.
  • They want to know which of their posts are pulling AdSense ads that generate clicks so that they can write similar content.
  • They want to see search trends -- which search terms from Google and Yahoo become more popular over time.
  • They want to learn which pages are the most popular during specific time ranges.

I can think of a few more things that might be interesting to correlate. I'm interested in knowing anything else that people might want to get out of their statistics.

In my opinion, it would be ideal if a statistics package tracked both WordPress and non-WordPress page requests, was as easy to get up and running as installing a new WordPress plugin, and presented essential data up front and in a more visual way whenever possible. It should allow at least some historical data to be recalled, and while it should not rely on a third party for that information, it should also not chew through server resources to store it. It should also be extensible so that when it doesn't do exactly what you want, you can easily add it.

Have any other ideas? Even if you're satisfied with the solution you're using now, it would be helpful for me to know what about that solution satisfies you. Because, as the good salesman says:

"Does your stats program satisfy you? Good, I'm glad it does, because my stats program isn't going to satisfy you; It's going to make you ecstatic."


39 Responses to Statistics, Anyone?

  1. Ozh from planetozh.com 1969-12-31 19:00:00

    The problem with Google Analytics, at least for me, is that it shows 100 complicated stuff when I need like 10 or 15. But once you figure out where to click to get what you need, this tool is hard to beat. It's also a perfect Adsense stat tool.

  2. AlexRaiano.com » Blog Archive » links for 2006-02-04 from www.alexraiano.com 1969-12-31 19:00:00

    [...] Statistics, Anyone? · Asymptomatic (tags: wordpress) [...]

  3. charlie from nfiniti.org 1969-12-31 19:00:00

    Actually I was looking for a nice (free) stats package as well. I currently use Awstats for some general info, I don't take it to heart too much. On the side, I've got a StatCounter account which I also use. I kind of compared the two.

    But reading the comments to your post, I though SlimStat definitely looked the most promising out of the lot and I just installed it on the server.

  4. Andreas from andreasviklund.com 1969-12-31 19:00:00

    PHPmyVisites.net - The only stats you will ever need. It is probably the best free statistics script in the world - atleast in the javascript-based category. It is complete, extremely detailed, professional and totally free. It comes with a great installer, and it supports multiple domains, multiple users (with different access settings for different sites), has IMPRESSIVE historical data, comes in lots of languages and draws a lot of pretty graphics for you. It also tracks outclicks, downloads, unique visits, returning visits and a lot of other things that I have not seen in any other free script.

    The only thing that PHPmyVisites does NOT do, is to display some kind of "overview" where the most important stats are shown at one single screen. But I have Mint as well, and Mint is a sweet and simplistic script which is great for quick overviews. But for the professional user who wants to ANALYZE the stats (and not just view it) PHPmyVisites outperforms Mint and every other script that I have ever tested. Try it out. I promise that you will not be disappointed!

  5. pippoppo 1969-12-31 19:00:00

    hi,
    i'd also suggest an italian project,
    http://www.phpstats.net/
    it's a php-mysql package, both in italian and english.
    in italy it's quite commonly used, don't know abroad.
    it's free for non commercial uses.

  6. Jeremy from www.axesandalleys.com 1969-12-31 19:00:00

    I was running AWStats for quite some time. Unfortunately it's notoriously a bitch and a half to install and run with my hosting company. Then it just suddenly pooped out on Boxing Day and showed only null stats. Then it just disappeared, throwing a 404. Then it reappeared, with more screwy or null stats. Then it disappeared again. The support forums (either through searching or asking) were absolutely no help, which seems to be the case for most needs as the majority of topics have no replies.

    I'm back to using my hosting-provided analyzer, which I believe inflates the uniques by at least 10%.

    I definitely prefer a log file analyzer and my prime desire would be to have a free one with real-time analysis.

  7. Owen from www.asymptomatic.net 1969-12-31 19:00:00

    Well, your cheap ass probably isn't going to get a path summary, but the rest of the stuff seems reasonably doable. RSS of stats seems like a great idea.

    The "And here is the one you've been looking for" moment hasn't arrived yet. When it does, you'll hear about it. :)

  8. Peter from blog.ftwr.co.uk 1969-12-31 19:00:00

    I am a big fan of webalizer and use it for all my stat monitoring.

    I have had to build a large list of referrers to exclude now due to all of the referrer spamming that goes on.

    Running webalizer on a hourly interval gives a good level of interactivity without burdening ever hit with updating the stats.

    apachetop is a good tool for live monitoring if you have shell access to your server.

  9. Erik Dasque from www.frenchguys.com 1969-12-31 19:00:00

    How about BlogBeat, it's blog centric, pretty simple and knows about wordpress.

  10. Owen from www.asymptomatic.net 1969-12-31 19:00:00

    @Erik: Please go back and read the post. The whole post. Then you can comment.

    Dammit I hate that.

  11. James from homedeco.geniosity.co.za 1969-12-31 19:00:00

    I use a combination of phpMyVisites and StatCounter for my blog. I like the summary pages of StatCounter, but I like the historical (and amount of) data that I get with phpMyVisites because it stores the data in a mySQL DB on my server... Of course, I don't really have a lot of visitors, so the DB isn't taking strain just yet. ;-)

    Nice review of the others. I was hoping for a case of "And here is the one you've been looking for". I need the following in a dream package:

    - RSS feed of statistics (call me crazy) - phpMV has this
    - Summary page showing Referal page -> Entry point -> Path thru site -> Exit page -> Destination page
    - Page AdSense ads were clicked (and user details from above)
    - FREE for my cheap ass

  12. jpep 1969-12-31 19:00:00

    I like mint, by from a maintenance perspective, it is a pain. It's tough staying on top of all the core and pepper updates. Also, I don't like that it doesn't have multiple users. Sometime I install it for clients, but I also want to review the stats on their site...

    Check our slimstat: http://wettone.com/code/slimstat

  13. Gnorb from www.gnorb.net 1969-12-31 19:00:00

    I've been using MyBlogLog (http://www.mybloglog.com) lately and I'm pretty happy with it. It's $25/year for the pro version, but they also have a free version which doesn't show as much info and has a 24 hour delay on the info. (I use the free.) It's proven to be a pretty good package, at least for my needs. You get a 3-day trial to their "pro" version when you first try it out, so you may want to go ahead and do that. The stats are kept on their server, not yours so that's either good or bad, depending on who you talk to and what you're looking for, and the only dent they make on your code is a javascript tag in your header file.

  14. Zoion from zoion.surprise-ending.com 1969-12-31 19:00:00

    I'm trying Delineate, it seems nice, but it's down a lot.

  15. James (aka MacManX) from www.macmanx.com 1969-12-31 19:00:00

    The "Trends" Pepper that Ian previously mentioned is great and should feed your hunger for "historic" statistics when dealing with Mint. Personally, I prefer to keep my "historic" statistics separate from my "live" statistics. For this reason, I use Mint to provide "live" and up-to-the-mintues statistics, while AWStats sits in the background and provides "historical" statistics by generating a full statistical analysis of my server's access logs once every twenty-four hours.

  16. Jason from jason.ungos.com 1969-12-31 19:00:00

    That's a nice run down of the different stats packages available. I've been on the search for a new stats program and have already used everything from free software (phpStats is one you didn't mention) to company-paid apps (WebTrends Enterprise software:non-hosted version).

    I really like WebTrends, but it's just not worth paying the money on a non-income producing site. You can configure just about any report or graph you want provided the information is in your log files. We even tied ours to a database which made on-demand reporting extremely fast. (Our logs were still txt files that were imported into the db). Bundle this with BrowserHawk, and you get everything a statistician would love.

  17. parrott from blog.parrott.com.au 1969-12-31 19:00:00

    I have used AWStats, Webalizer and find AW the best. But what I would like to know is if anyone has an idea of a analysis tool that helps monitor subdomains as well. Tracking www is fine but when they agregate all stats under www it is hopeless.

  18. Ian D from www.iand.net 1969-12-31 19:00:00

    I use Mint and have the same issues - hard to see past the last few hours/days. One pepper that helps though is which allows you to compare one period against another i.e. the last 28 days against the previous 28 days.

  19. Greg 1969-12-31 19:00:00

    I use webalizer, but have experience with Sawmill as well - if you configure your webserver logs to capture stuff like resolution, Sawmill will let you do all kinds of queries.

    It's not the most user friendly of programs, and it really grinds away if you ask it to do too much though.

  20. Fabio from www.montescudaio.com 1969-12-31 19:00:00

    I have been using some of the tools you mention, as well. Stattraq is nice but for me it has two problems. It is limited to wordpress pages and it is not able to understand if a request comes from a known bot/spammer of it is a real user. It did not seem actively developed last time I checked.

    Later I switched to Analog (a log analyzer program) nice and very configurable but hard to use. It had the bot/ spammer confusion, too. I had to change apache to resolve the IP of every visitor to get meaningful results and this put too much load on my server.

    I tried Shinistat but it is way too limited in the free version, and so I was happy when Google Analytics was released. It does count only meaningful result and has ton of stats. Too many options actually (paricularly if Adsense is not used, like in my site). But when you understand where the data are it becomes more manageable and some of the stats are really helpful and not available in other script based services.

  21. Jauhari from nurtria.net 1969-12-31 19:00:00

    I just use Google Analitycs and Mint for this time

    This really great tools :)

  22. Sam from www.diyhappy.com 1969-12-31 19:00:00

    I personally use ninjadigital for my stats. They have a great interface and tons of features. It's very intuitive. Their site is www.ninjadigital.com

    Although it's not a free solution.

  23. vitanigma from japanation.de 1969-12-31 19:00:00

    You've kept your promise. Thats truly a good hint Andreas. I tried it out and was surprised: all the Information I need is given and the tool was easy to handle. It even supports newsletter and partnership.

    The bad thing, even for PHPmyVisites, still is: no tool, at least none I know, can handle "old" logs. In my case that is webalizer wich was in use for many years now. Would I be switching this means a loss of all historical data.
    Are there any workarounds somebody knows?

  24. HardinComp from hsir.technosailor.com 1969-12-31 19:00:00

    I use a pair of stats packages regularly. The first is your BAStats, which works really well for seeing what's happening. The other is TraceWatch, which I've been able to implement site wide with no trouble. (I also use it on my forums and online store, which are all on the same server.) TW gives me not only some fantastic data, but also the ability to drill down to see a lot of detail. The only thing it lacks, other than the ability to see page titles for my blog pages, is the ability to see which link was used to exit my site. I recommend this program to all of my clients as well as to several friends who I don't do design work for.

  25. drac from lair.fierydragon.org 1969-12-31 19:00:00

    One thing that I miss in all the stats packages you've described (yes, tried most of them) is the ability to drill down to more detail if necessary.

    Yes, the general use case is for overviews and Shortstat serves my purpose there. But Shortstat _only_ shows the overview and it's impossible to find out or even link the "who" or "where from" behind a specific request or series of hits if I needed to.

    I suppose another crucial part of the puzzle is outgoing links and so on - but there is an essential tradeoff to be made by using Javascript (and having access to nifty information like screen res) and using a query class rewriter like Shortstat or StatTraq. I don't like Mint and Google Analytics because they don't track people who either block the .js or don't have Javascript enabled anyway.

    I'm still mulling over that idea about writing a different presentation frontend for the data collected by Shortstat :)

  26. Simon from www.photogabble.co.uk 1969-12-31 19:00:00

    It would be nice to have a stats plugin for wordpress that simply produces a nice graph outlining the visits and page views for the past 12 months and/or detailing the months view as well as the top say 12 or so popular articles on a blog. This plugin could be included in a page such as stats, so it is part of your website and stuff...

    If I cant find such a plugin I will write one myself :) I use awstats and webalizer (although I prefer awstats as it is more comprehensive). But awstats nature means that the top 10 pages that it sees as being the most popular are infact .css files, and ofcource they are going to be popular, they get hit everytime a page loads :/

  27. Emily from How to Blog from www.emilyrobbins.com 1969-12-31 19:00:00

    Hi Owen,

    Have you ever checked out Sawmill stats? It's the best and most comprehensive log analyzer I've ever used and I find it provides me with all the data that I need for both stats and SEO efforts.

    You can set it up to tell you what the keywords/keypharases and referrers were for each particular page, which pages on your site have been spidered and when, and much more. I've been using it for over a year now on all of my sites that are hosted on my own server, and I LOVE it!

    Hope that helps!
    Emily

  28. BrittaBlog » Discussing stats packages from britta.wordpress.com 1969-12-31 19:00:00

    [...] Statistics, Anyone? [...]

  29. Goodspeed from www.stanley-goodspeed.com 1969-12-31 19:00:00

    Hello everybody,

    after I installed stattraq und wp2.0, my rss-feeds doesn`t work anymore. Has anyone the solution for my problem?

    Thx.
    Greets from Germany
    Goodspeed

  30. Psi from www.psicotico.com 1969-12-31 19:00:00

    In a small traffic site ( http://bbclone.de/
    I doesn´t need a MySQL database and has realtime statistics with referes and even resolves IP. You can even use it on a non PHP page forcing Apache to read a html file as php.
    A very handy tool and very easy to install.

  31. Joe Tan from tantannoodles.com 1969-12-31 19:00:00

    I've created a WordPress plugin which pulls in Google Analytics data (and feedburner), and displays it in your WordPress admin screen.

    http://tantannoodles.com/toolkit/wordpress-reports/

    It's still a work in progress, so I would welcome any suggestions you have. I'm also hoping to add a trends report soon.

  32. andy from www.yellowswordfish.com 1969-12-31 19:00:00

    Well actually Owen, I kind of like BAStats :) but sadly the author didn't seem to get it finished and it looks like he has abandoned it. Mine still says 'beta' at any rate. Oh yes - and the purge doesn't work!

    But a new, updated BAStats with a few more functions and options would be just great.

    Any chance?

  33. Jason from jasonbainbridge.com 1969-12-31 19:00:00

    I swear I skim read through the article and even tried a search for measure before posting that last comment but then again I did do it before my first cup of coffee so please excuse me while I go hide in the corner...

  34. Evan from ebroder.net 1969-12-31 19:00:00

    I've been using BAStats as well, and I like it for the most part, although the fact that it's been abandoned has encouraged me to look for an alternative.

    Of course, I don't try to make any money off my blog - it's just for fun, and I only keep stats to see who comes to visit my site, so what I want in a stats package is different from what others want.

  35. MacManX.com » Blogroll Dive: 2/6/06 from www.macmanx.com 1969-12-31 19:00:00

    [...] Owen discussed the pros and cons of several web statistics packages. [...]

  36. Jason from jasonbainbridge.com 1969-12-31 19:00:00

    What about MeasureMap? It is still in closed beta but if you sign up you should get an invite fairly quickly.

    I haven't used it very much yet myself but I think it really does give you the information you want on the front page and allow you to drill down from there, which is exactly how a statistics program should work.

  37. Mani Monajjemi from mani.wordpress.com 1969-12-31 19:00:00

    I used statTraq for about 6 months in my persian blog (www.manionline.org/weblog) , You are right, a very huge database will be created , My Database (before statTraq uninstall) was 6MB large!

  38. Richa Avasthi from blogs.ravasthi.name 1969-12-31 19:00:00

    I'm really surprised no one has mentioned SiteMeter yet. Unless I'm missing something?

    SiteMeter seems like (other than the fact that it's a 3rd-party solution) it's got most of the features you say you're looking for: stats for visits and page views; tracking by day, week, month, year; all the JavaScript goodies like monitor resolution and color depth; entry and exit pages; and my favorite, viewing recent visitors on a world map. Plus it's free.

  39. PetitOn » Blog Statistics from petitpub.com 1969-12-31 19:00:00

    [...] Owen at Asymptomatic has tested quite a few statistics packages in an effort to get some clue of who is reading his blog. And as he was at it, he also kindly has given his view on the pros and cons of the different software and services. The post has already drawn some attention and many comments on the solutions he discusses, as well as suggestions on other software and services. [...]

489