There are various terms used to describe the science of recording and interpreting website statistics. Web metrics, web analytics, web stats and site stats, to name a few. 'E-metrics' refers to analysis of electronic businesses. Here is a guide to the industry terms and what they mean to you and your online business.
The 'metrics' of web metrics refers to measurement - the science of measuring websites. Specifically, measuring website “events”, and extracting trends. In this case, those “events” are human clicks.
Analytics is the act of distinguishing categories within recorded stats, and analysing for any patterns.
Statistics are a scientific application. The goal is to form actions, for example website content management, based on the data which are recorded.
With an application of statistics there is less guesswork. Simple questions can be answered, for example, something very basic:
- Are there more or less people coming to your site this week than there were last week?
- Is your site doing better or worse this week?
They will inform you about numerous aspects of your traffic; the number of (returning) visitors to your site, and how visitors surf through your pages. This information tells you about the content of your site and how visitors use it. Your traffic statistics are an indicator of website performance. When applied in this sense, site stats can be used very effectively to make updates.
When comparing different types of measurement the classic scenario of 'the difference between apples and oranges' often arises. In the same way, different website statistics programs have unique ways of measuring.
Therefore it is not always easy to compare the results generated by two stats programmes to track one site. The process itself can be very useful, in terms of thinking through the differences in results, and determining what is actually being measured. We encourage the use of numerous programmes, for example, combining a tracking service with log analysis.
If the method of measurement stays the same through time, then the results will be very useful for purposes of comparison. Therefore, choosing the method of measurement is important.
If you compare results from two types of measurement you will find differences in numbers. If you compare the same statistic over time, however, then you are not changing the method of measurement. This is the most accurate way of recording statistics. This will allow you to find patterns and definitive answers, for instance if traffic is growing or diminishing.
In any statistical endeavour, the first step is to define what is being measured. In this case, the common denominator is human events, clicks on a website, which are defined as page views.
Specifically, the statistics discussed here are a translation from raw data, clicks, and server-browser dialogues, into a user interface from which patterns can be discerned. The goal of web metrics is to extract patterns which tell you what is happening. The next step is to create actions, i.e. what to do about your traffic patterns.
Web metrics and web analytics is an exciting field at this moment, because there are not many patterns being sought. An example might be comparing 'bounce rate for first time visitors' with 'bounce rate for returning visitors', which has not become a standard of analysis (aggregate bounce rate stats tell you how far into your site visitors are clicking).
Nothing can be measured with 100 per cent accuracy, this is not the skill that we endeavour to reach. The skill lies in trying to keep measurements useful, despite the inability to reach 100 per cent accuracy. An acceptable margin of inaccuracy is one per cent. That does not make the world an uncertain place; it means that you have to be specific in knowing what is important. For example are trends rising or declining over time?
The process of determining what to measure involves the creation of numerous definitions. There are always elements that are being under or over-measured. That is why the system requires constant calibration, in terms of what people really want to know, which in turn determines what should be measured.
Statistical needs vary depending on site size. Therefore it is up to statistics programmes to present the statistics in a way that is useful for webmasters of different sized sites.
Large sites for example, are more interested in trends. Due to the volume of data a larger site generates, a single clickstream (or click path) won’t be very interesting. As there are too many clickstreams (e.g. sites which receive several thousand visitors a day), only the aggregates are interesting. Larger sites are interested in aggregate data, while smaller sites are interested in discreet data.
Trends are aggregate statistics. For example, a site's bounce rate is an aggregate statistic. Bounce rates are stats designed for the purpose of identifying patterns which are hidden within the stats.
Discreet stats such as clickstreams, will tell you what individual people are doing on your site. Discreet stats are not aggregates, as you are actually seeing what the data is "built" of.
This type of information clickstream analysis is very useful for development purposes and understanding user reaction. If you are designing a new site, knowing how first-time visitors navigate will help to determine how successful the site is, and what changes need to be made.
To illustrate some of the difficulties associated with counting and measuring, consider a statistic that tells you how many people voted in an election. Counting votes is a difficult process and re-counts are often undertaken and it’s not unusual to reach different totals every time.
When polls are released, the number presented is an extrapolation, based on a percentage of people contacted by phone, or asked at the door for whom they voted.
I believe in presenting trends derived from actual clicks, which will narrow the margin for error. Optimisation techniques based on cookies and visitors improve accuracy.
These illustrate page views over time, the traffic is deduced from unique visitors and there is minimal 'double-counting' of visitors. I am often asked why Opentracker’s traffic numbers are often lower than those recorded by log files – and this is why!
The important thing to remember is that data, (i.e. statistics) are numbers created by people. Therefore it is crucial to understand how these numbers are defined and generated.
Eddie Moojen is co-founder of Opentracker and creator of the statistical engine on which its services are based. He is a regular
industry spokesperson on the issues surrounding website statistics.
www.opentracker.net
