Web Analytics Works Shock!

Web Analytics, or the science of Guesswork (subtitled We take the Guesswork out of Heuristics).

Having struggled with Webtrends for more years than I care to remember, I have a healthy disrespect for the Snake Oil known as Web Analytics.

Two problems really spring to mind.

Firstly the difficulty of trying to work out what happened by looking at log files where you never know the ending. The stateless nature of HTTP means that you never know when someone has gone. [aside, I was going to put a neat link on stateless to a good definition, but I can't find one so...... Stated Communication is like a phone call where you dial a number and a connection is made with the far end, you can communicate over that connection until it is terminated, Stateless Communication is really impersonal you dial the number with a question, it responds and forgets about you, you call again and it doesn't know who you are!] so all this bollocks about "average time on site 21 minutes..." is just that, bollocks because you never know when someone has gone.

The second is trying to display the log files in a way that is understandable to man. I mean you've got a load of garbage data in.......

However, we recently started running WebDruid (which I think is an offshoot of Webalizer) which does seem to work. And has the most intuitive "Paths Through Site" diagram that I have ever seen.

I once sat through a meeting with a large corporate client trying to explain Webtrend's Paths Through Site diagram, a complete disaster. Even Webtrends can't explain what the diagram is, so mere mortals like myself are just lost.

Anyway, we were looking through the logs to one of our own sites, Enabler our email marketing software and we noticed that, we were driving traffic to the Contact Us page which we could see clearly from the diagram but were't getting many enquiries. Strange, we thought. But looking at the page we realised that Contact Us had no email address to contact us by (grim for an email marketing product!).

Fixed it now, lets see if we get more enquiries.

So the sequence of events was, looked at the web analytics, noticed an anomaly, checked it out and found a problem, fixed problem.

I should do more of this for both us and our clients.

ID Cards part n

The subject of ID cards in the UK (I live here!) is one that continues to disturd me. I am in opposed on both principle and in practice, and concerned about the motives of those who are pushing so hard for them, oh and it doesn't work (thank you Tom Wolfe).

My opposition in principle is because I don't believe that it should be neccesary to continually "prove" my identity to anyone who asks for it. I am my identity, that's it, end of story. An ID card is simply a thing that may or may not confirm that at a certain point in time there was a correlation between me and that thing. It doesn't prove anything. And as soon as you have an ID thing, you of course have to have the immense central database to support it, as well as the freely distributed "checking terminals" that have access to it. That is scary, oh and it doesn't work.

In practice, something that I am sure that I will expand on given time, there are some real, I was going to say difficulties but really they are show stoppers, issues. One I shall touch on today is this that of easily accessible technology.

One of the problems as I see it is that any identity card is going to be based on some form of mass market technology, e.g. some piece of plastic in roughly a credit card size and shape that has embedded a "chip". Now, I don't expect the UK Government and their agents to invent a whole new and unique way of doing this. That is, to invent a new technology, packaging technology and read/write mechanism et. etc. etc. Oh no, they will use off the shelf components and technologies and who wouldn't? And as soon as they do that (or soon afterwards) it doesn't work.

I can still remember in the 1980's getting my first mag stripe writer (from a project that I worked on) and all of a sudden I was able, simply and without writing any code to clone credit cards! Unbelievable. All I needed was about £200 work of kit and away I could go. Obviously I didn't otherwise I might be posting this from inside Wandsworth Prison not from outside it.

Compare that to the effort and specialist skills and a fuck off printing press and finishing equipment and so on that I would have needed to copy a cheque book.

Today my "official identity" probably consists of documents e.g. Passport, Driving Licence, Birth Certificate etc and snippetes of information held in lots of unconnected databases like is National Insurance (Social Security) number, tax records, employment records, Electoral Register, the list goes on. All of which would have to be created/updated, many of which are pretty closed off.

The bottom line is that to do a good job of creating an identity today you need a lot of time, effort and access to some pretty specialised equipment, with ID cards, you just need to be able to copy one pretty standard technology ID card and update a database that has tens if not hundreds of thousands of terminals attached to it. I know which one I think is easier!

So in a sense ID cards make us less sure of confirmed identity, whilst having the appearance of being more sure due to the "infallable" ID card, oh and it doesn't work.

Silly Observation

Two of the regular events at my office have swapped over recently.

We have a daily visit from a sandwich delivery service, which is getting earlier and earlier. Today it was 10:50, somewhat early it must be said. The speculation here is that if they can get everyone to eat lunch by 11:30, then there is scope for a second delivery mid afternoon. It hasn't happened yet, but it may.

The other event is our mail delivery. This as opposed to the sandwiches is getting later and later.

And we have now reached the point where sandwiches can be delivered before the mail, which is bollocks. No wonder the Royal Mail (for it is they) can make £550m profit, they don't deliver the mail at a resonable time.

Test Post

This initial Test Post was posted from London on 24th May 2005