The IP Development Network Blog

About a month ago, I undertook a small project to evaluate a couple of visitor stats packages so that I could get a better idea of what readers wanted from this blog. I am not going to produce anything as wide ranging as can be found on conversionrater.com, but seeing as I went to the trouble of comparing a few of the bigger ones back to back, it is worthwhile summarising my findings.

Quite what use this will be to you (unless you too write a blog), I have no idea. This might just be an example of how you get what you pay for (nothing), but I will try and throw in some wider observations to keep you entertained along the way, even if you don't much care for stats. You might be interested in how much information about you is left every time you go through a site...

I started the blog with partial feeds - giving prospective readers the chance to choose whether to read on. A click through is then a hit and I know I should write more like that, but it puts a lot of emphasis on writing a good headline and succinct opening statement.

I started off with StatCounter. Like all the other options, I have to install a small amount of code on your web page which then contacts the StatCounter server and tells it your IP address, the page you are accessing, the page you came from and any internal hyperlinks you click through on the page. It also captures the times of all these events and stores all this information in a log file.

The size of this log file is determined by how much you pay them. You buy a certain limit which is then stored on a rolling basis, dropping the oldest record to store the next one. You used to get 100 lines for free, but this has just gone up to 500 lines stored as a way of hooking you into their service. The problem is that the rate for additional storage goes up quite quickly after that, although this has also been improved in the recent package upgrades.

StatCounter doesn't just give you the logs, although these are also available to you if you want to process the data yourself. One of the good bits about StatCounter is that it gives you a predefined set of queries that you can really use and tell you who your visitors are, where they are coming from and where they are going to. It is in the depth of readily available information about individual visitors - the drill-down - that StatCounter excels. The Visitor Paths query shows me your IP address, your network provider, the page you came from and any search strings you brought with you, the pages you read on this site and any links clicked through to other pages on this site. Unlike a few of the other services though, StatCounter does not register clicks to links on my pages, which take you off this site.

One of two individuals even leave their name in the logs, presumably because they use a fixed IP which is recorded in their name. Instead of a network provider's name, I can actually see the name of the individual. I'm sure Martin Geddes won't mind me picking on him as an example. Scary, isn't it...?

It was therefore no surprise when my readers started asking for full post feeds. It is a lot more convenient for the reader and it hides their identity behind a Bloglines account number or the equivalent. A lot of readers like to lurk - it's human nature - I can understand that, but I wonder if they consider the even greater level of information they give Bloglines or whoever about who they are and what they like across a multitude of sites. I'll wager a small side bet that these reader services get very good, very quickly at targeted advertising...

Anyway, so I went to full post feeds but I wanted to keep track of the articles that people were reading so that I could write more of the stuff they wanted and less of what they didn't. With full post feeds, readers don't need to come by the site so you don't get a hit on StatCounter. The log files are useless: the number of requests for the RSS feed is misleading because aggregator poll your server every so often counting as a hit whether or not you deliver any new content.

So I set about trying to use Feedburner.

I quickly realised that the level of basic information about my feed in the free service is not enough. That just tells me how many subscribers I have, not how many of my articles get read. For me, this was like a chocolate teapot (useless), so I had to upgrade to the Total Stats Pro service where this information is available.

Even so, I am not in love with the level of information you get, so now that I am paying $4.99 a month, I am questioning it. The problem I have is that although I like the StatCounter (and Google) services much more, Feedburner is the only service I have come across allows me any view at all of how many feeds get read. Even so, there is no way of knowing how many feeds get "read" by clicking "mark all as read" versus how many are actually read (nor do I get the time taken to read that you get if the pages are on your own server).

I have considered removing Feedburner, but doing so would mean either a return to partial feeds or giving up any view of the relevance of the content that I am producing for regular subscribers. It strikes me though that full vs partial feeds is one of the major differences between old new media (The Register) and new new media (GigaOm). With old new media, they are the aggregator. With new new media, you are. Purely for audience tracking purposes, headlines, snippets and click throughs would certainly be my way of choice, but this is new new media, so your full post feeds are safe.

Feedburner also give you the standard web site analytics option, but to be honest I don't use this very much because it is second rate compared to the others in the sample. The interface not nice looking, and not user friendly. But it is free, but then so is Google's Analytics...

I will probably hold off cancelling Feedburner until Google add the ability to track RSS feeds to their otherwise excellent package. I'm sure it will come, it is the one big hole in what they offer.

I signed up to Google just before the enhancements to what is still a Beta package but I already have a hugely impressive array of reports. The USP here are certainly the graphs and the look, which are lovely. Soft on the eye and really informative - very 2.0 chic. If you are an artsy type, Google is definitely for you.

The dashboard is really very good and affords the ability to drill down into the same level of detail as StatCounter, albeit without quite the same clarity of intent. With Google I still find myself somewhere and think, "oh yes, that's interesting", without being entirely clear on what slice of data I am looking at or how I got there.

There is one big advantage for Google over StatCounter. Google is free no matter how big you are, as far as I can see... They have yet to ask me for a penny, but then this site is just a tiddler so I don't know for sure what they do if you hammer them, but there doesn't seem to be a limit on records stored.

I'm sure Google makes money somewhere, but as with all Google services, the user doesn't pay - the advertiser does. This is how Google is making money from analytics. The service not only provides the hit reports as do the others but it allows you to build in goals (benchmarks) and track how you are doing. It hooks you in and allows you to hook in ad-word campaigns. Clearly Google sees analytics and a way to get people thinking about ad-words when they are thinking about how to get more visitors. Stunningly simple, but probably very effective.

The other one I tried was Clicky. This is another one with a very nice looking interface. In fact it is probably even more 2.0 than Google, but Clicky doesn't seem to have anywhere near the same summary to drill down capabilities that Google or StatCounter have. One upside is that Clicky "integrates" with Feedburner, essentially allowing you to suck your Feedburner data into the Clicky interface.

I have however, found this to be a better idea than an implementation as there seems to be a number of design and technology issues there. For example, the Item Views that I get with Total Stats Pro from Feedburner seem not appear reliably in Clicky and the view of the Feedburner data is very one dimensional without drill down capabilities. The records I can analyse are only available one day at a time...

So where does that leave me? If I was just going for one interface, I would have to for the all-round averageness of the Clicky product, or the somehow unsatisfactory Feedburner service. This is because, as I have already mentioned, I need to track the full post feeds that I issue.

It is however much better to be using all of them in combination right now because you can draw more conclusions (and iron out the many slightly contradictory messages) if you can invest the time to look at all the data side by side. This is because none of the services meets my needs. My ideal would have the Google Dashboard and log size constraint (none), the StatCounter drill down and (a better version of) Feedburner's read item stats.

Greedy aren't I...?

Labels: blog analytics, Clicky, feedburner, Google Analytics, StatCounter, targeted ads

# posted by Jeremy Penston @ 6/01/2007 03:30:00 PM

Friday, 1 June 2007

Visitors