Matt Hopkins: Web Analytics Consulting & Online Marketing

Web Analytics Methodology Part 20
Matt Hopkins posted in Web Analytics on May 20th, 2007

In the previous part we dealt with the way of describing KPIs (Key Performance Indicators) but each website is unique and so requires different KPIs. A KPI is a way of measuring the success of your website and so is based upon the tracking metrics of your site analytics.

Lets take the example of this blog. I measure the number of visits, unique visitors, RSS subscribers, internal search keywords, external keywords and referrers. From this data I can calculate the number of visits via my RSS feed as another metric and a KPI as the number of RSS visits per RSS subscriber.

The main purpose of my site is to provide educational information to people interested in web analytics, to achieve this purpose my main objective is to increase the readership of my site. So my new KPI will effectively let me know how one aspect of my readership is doing.

So, I have identified my site purpose, objective to achieve my site purpose, a method for measuring my site objective and based a KPI on this method. Over time I would hope that my KPI of RSS readership per subscriber will increase.

How Many KPIs Do I Need?

How many KPIs can you handle? 5? 6? 20? It is said that people can only handle a maximum of 7 pieces of information at any one time, so with this in mind I would suggest no more than 7 KPIs on a dashboard per person.

You should use the rule that if you have a KPI that someone could not make a direct business decision based upon then that KPI is useless and should not exist.

Who Needs KPIs?

The only people that need to know the value of KPIs are those people that can act upon them, this is reiterated in the table from the previous part of this methodology.

What is a Tracking Plan?

To be honest this is just a glorified name for taking the above table and populating it. To achieve a goal you normally need a plan to get there, be it in business, in life or with your website. This table is my version of a plan for achieving your website objectives. The idea being that you fill it out once when you design your website, but it is not just left, instead it should be a living document that grows and changes just as your business changes. Set a quarterly review cycle for looking at the table and seeing if the KPIs set are still relevant to your business, or has the market/technologies changed in a way that you need to meausre success in a different way. For instance if you initiate an RSS feed on your site then you will need further KPIs to meausre the success of this RSS feed.

Part 3 will cover capturing your web analytics data, picking the right tool and hiring a web analyst.

Catch up on other parts of the web analytics methodology and subscribe to my RSS feed by clicking the orange logo at the top right hand side of this window.

Summary: Web Analytics Wednesday - London 9th May 20070
Matt Hopkins posted in Web Analytics on May 13th, 2007

We hosted the second Web analytics Wednesday in London of 2007 and based on the feedback from the previous event we made some minor changes.

Firstly, it was hosted in the trendy Leicester Sq. Ruby Blue bar which was a great improvement in my eyes as we had our own room with its own bar. Unfortunately it was raining so the numbers appeared to be a little down on the last event despite more people registering this time, but alas we can´t control the weather.

We started the night with a short presentation followed by a disucssion based forum around the subject of Search Engine Optimisation using web analytics. Myself and one of my colleagues (Bob) hosted the presentation and started things rolling with the discussion. However I don´t think the audience expected to have to participate as it seemed a bit of a struggle for them to speak, although I did have some valuable asisstance from
Rufus at Clickstream. This was a whole new area for me as I´ve never done a ´public´ presentation before and there is a lot to learn for me from this event about getting people in the right mood for talking and how to read non-verbal feedback from the audience.

But anyway we will take inspiration from this event and hopefully make the next presentation even more engaging, if you were there and have any suggestions for me (or Bob) then please leave a comment on this post.

Unfortunately I had to leave straight after the presentation due to personal commitments but i am assured that the rest of the evening went well with some interesting people turning up like web analysts from ITV, BT and Channel 4.

So what were people talking about in the discussion?
Well, there was mention of the lack of tracking ability of web 2.0 technologies such as podcasting and videos. Someone mentioned the Comscore cookie deletion rates, personally I think this topic has been done to death and people just need to realise that web analytics is not an exact science and probably never will be.

There was a debate about what the best type of web analytics was, should you investigate erroneous data points in your graphs, just report the data or simply test different campaigns using A/B or multivariate testing. Our argument was that you should do all three but I don´t think there was any sort of concensus with that one.

That´s about all for this WAW summary but if you want some info on when the next one is then go to the SCL Analytics website.

How To Clean Your Web Server Logs0
Matt Hopkins posted in Web Log Files, Web Analytics on May 2nd, 2007

One of the main differences between using web server logs and using page tags as your data collection method for web analytics is that robots and spiders are tracked within web logs but not in page tags.

Now 99% of people looking at their web analytics will want to concentrate on analysing visitor traffic rather than what the robots and spiders are doing and looking at. Given this it is important when analysing web server logs that you accurately identify robot and spider activity. With this in mind I have written a short checklist of things to look out for that may indicate a robot or spider.

• User agent, this is the browser that is displayed in your Browser report. Well behaved robots and spiders will identify themselves within their user agent string and as such they can be quickly identified and removed.
• Visitor duration, this can be seen in a visitor report along with the total time online metric. Given this informatoin you should also be able to see for one particular day whether a visitor has viewed over 100 pages in 1 or 2 visits or maybe spent all day on your website. These kinds of behaviour are common of robots and spiders and as such should be removed.

Other ways in which you can clean data represented in your web server logs are listed below:

• Excluding internal IP address, the aim of this is to exclude any internal traffic that may be accessing the website as this would scew the visitor behaviour. If you are running on dynamic IP addresses then you may want to think about tainting your browser’s user agent string to include your company name then exclude internal traffic by user agent. A quick search on google for modifying your user agent should point you in the right direction for this.
• Excluding irrelevant pages, if you have pages live favicon.ico and robots.txt or your /admin/ directory then you may wish to to exclude this data as there will be lots of requests for these pages but they are more like web site resources rather than requested page views. Unless you are monitoring your admin area usage, this data will not be of use to you.

The final thing I am going to mention is really that an analytics package can only ever report on the data that it is given. So if you page report looks a bit cryptic then you may want to consider rewriting your URLs in a user friendly way rather than using dynamic URLs. Also making data available via the querystring in the URL will help when it comes to analysing your web server logs.

If you feel that I have missed any other important items that should be removed from your web logs then please post a comment and let me know.

Thanks

How To Accurately Track RSS Subscribers7
Matt Hopkins posted in Web Analytics, Search Engine Marketing, RSS on May 1st, 2007

That title may be a bit over ambitious, but I think that I have come up with the most accurate method for counting subscribers to your RSS feed.

This post is the sequal to Tracking your email campaigns and will build upon the tracking concepts used within that post. The previous post on tracking allows you to identify visitors who read the RSS feed then visit your site through a link in the feed. So that method will go some way to providing a readership metrics for our RSS content.

A readership metric is great, what about subscribers?

We can track the number of subscribers to a RSS feed in a number of ways, but after a few hours of research one method stuck out. This was to taint the feed URL for every visitor with a random value. Then all you have to do is count up the values and that will give you a subscriber count.

Below is an example feed URL that has been tainted.

http://www.webanalyticmatt.com/feed/?ID=1234

In this case we could count the number of ID values to give a subscriber count. Now this is fine but my method is to take this one step further and have the random value as the persistant cookie value for that visitor. By doing this we can actually identify the visitor even more accurately than using cookies because the feed request becomes the visitor identifier. Therefore it is possible to constantly update the visitor persistant cookie, even if it has been deleted to reflect the original cookie value.

Obviously this method is not without its problems, but luckily the negative effects can be minimised.

For instance you cannot let the search engines get hold of the RSS feed URL as search engines seem to favour RSS content over native web content. To ensure this, make sure your robots.txt file denies access to your RSS feeds.

What if someone syndicates your RSS feed on their site? Well your readership metric should hopefully increase but your subscriber metric will not be changed. Not a great thing but I think it’s an acceptable risk.

No more pinging your RSS feeds to RSS search engines. This would obviously defeat the point of counting of counting subscribers as you would be promoting one subscriber ID which may actually represent many people. But if your content is any good and your search engine optimisation is up to scratch then you shouldn’t need to tell the RSS search engines where you are.

Now the finale, I have read a few blog posts on tracking RSS subscriber numbers but the one thing that differentiates this post is that I am confident enough to implement it on my own blog as well as any site I am working on. To see, just subscribe to my RSS feed.