There is often confusion over what exactly is meant by hits, page impressions, unique users etc.

Below is a guide to some of those metrics that can be used for measuring a web site’s performance with comments on how to get the metric, its advantages and disadvantages. Values can be assigned to each of these metrics and used as bargaining levers in any commercial partnership or advertising negotiations. You can’t value what you can’t measure…

1. Registered Users

Description
A user who has entered their details on a site, usually as part of a log-in / registration process. When they log-in to the site (usually using a username / password combination) they are identified by the server and their details known to the system.

How do you get this info?
You have a database which captures the users’ information. You can then mine the data as required.

Pros
By far the most accurate of user identification, tracking and data capture techniques as the user identifies himself / herself to the server. It does not matter where the user is or which computer they are using. The potential for user profiling and analysis is much greater. No. of registered users has been used as a way to determine the value of an e-business based on the perceived lifetime value of each registered user.

Cons
Do you really want people to have to register and log-in? How do you administer forgotten passwords etc. You might get the same person setting up multiple accounts / registrations. You have to take security and data protection very seriously if you start to store extensive personal information, in particular credit card details.

2. Unique User

Description
A unique user is an individual user determined by the user’s IP address or cookie.

IP address = Internet Protocol address identifying a computer connected to the Internet.

Cookie = persistent client-state HTTP cookies are files containing information about visitors to a web site (e.g., user name and preferences). This information is provided by the user during the first visit to a Web server. The server records this information in a text file and stores this file on the visitor's hard drive. When the visitor accesses the same web site again, the server looks for the cookie and configures itself based on the information provided.

Unique user figures aim to show how many different people have visited a site (over a month usually). They do not tell you how many times each person visited, how long they spent at the site, what they looked at etc. They have become an increasingly popular currency for measuring site reach in recent times – New Media Age now ranks sites according to unique users instead of by page impressions as it used to do.

How do you get this info?
The IP address of each site visitor is stored in the web server’s log files. Web servers (like IIS, Apache etc.) write log files to the server’s hard drive which detail site activity. A typical line from a common log file would look like this:

10.0.1.10 - - [05/Oct/1999:11:00:26 +0100] "GET /apache_pb.gif HTTP/1.0" 200 232

You can see the IP address at the beginning of the log file. You can do a ‘reverse DNS (Domain Name Server) look up’ to find out which domain the IP address came from. Log file analysis software such as Webtrends does this for you.

Pros
Unique users is becoming a standard metric to demonstrate the reach of web site.

Cons
Unique user figures can be heavily skewed both positively and negatively. Most corporate internet users connect to the internet through the company proxy server. Proxy servers can control and monitor access and cache frequently requested pages locally to conserve bandwidth usage. However, because of the proxy server (which has one IP address), all users will appear as only one unique user as far as the log files are concerned. This means that unique users will be underestimated.

On the flip side, most home users will go through an ISP to connect to the internet (dial up connection). ISPs will typically have fewer IP address that they can use (there are a finite number of IP addresses available) than users, on the basis that it is highly unlikely that all users will be connected at once. They allocate the IP addresses they have dynamically as users connect and disconnect. This means that one unique person is likely to appear as many different IP addresses in the server log files. This means unique user figures are overestimated.

Working out a weighting to come up with a fair figure would depend on the extend of home / work usage of a particular site. In practice, this weighting is rarely applied and the positives are considered to balance out the negatives i.e. unique user figures are taken at face value from the log files.

3. User session

Description
A session of activity (all hits) for one user of a web site. By default, a user session is terminated when a user is inactive for more than 30 minutes.

How do you get this info?
From the log files (see above for info)

Pros
This metric gives a good indication of how many visits a site has had. It doesn’t try to indicate how many different users have visited the site or how many pages impressions each visit created. It can be used to calculate browse to buy ratios = how many visits (user sessions) / how many purchases.

Cons
The user session figures can be massaged by changing the default session termination duration e.g. if you set the duration to 5 mins you would get a lot more user sessions than if it was set to 2 hours. This duration can be changed either in the web server’s session timeout settings or in the configuration of your log analysis software.

4. Average User Session Length

Description
Rather surprisingly… the average length of a user session. Anything above 10mins is considered pretty good going.

How do you get this info?
Good old log files analysis.

Pros
Gives you an idea of how ‘sticky’ and ‘compelling’ your site content is. Or how long the pages take to download…

Cons
Follow on effects of unique user error margins. There are also some non-human visits (e.g. search engine spiders) which could be argued to drag down your average user session length if they account for any sizeable percentage of your traffic – you’d have to have quite a small amount of traffic…

5. Average page views per user session

Description
Take the total page views and divide them by the user session. Hey presto. Around 15 page views per user session is a rough benchmark. The more newsy and the bigger the site (BBC a good example) the higher you would expect this figure to be. If you’re a web publisher relying on ad revenue you’d want this figure to be as high as possible.

How do you get this info?
You guessed it, the log files.

Pros
As more figures become available this could become a useful benchmark for comparing the page-impression-generating success rates for ad revenue driven business models.

Cons
None that spring to mind

6. Page impressions / views

Description
A page impression is logged as a request for an HTML file to the web server.

Page impressions are the common currency traded between web publishers and advertisers. Page impressions are sold by the 1,000 and costed at a CPM rate = cost per thousand page impressions. CPM rates average around $15 in the US and £20 in the UK. The UK rate is expected to fall as supply begins to meet the demand in line with what has happened in the US.

Banner ads are usually sold in 3 standard categories at varying CPM rates: Network, Channel and Single Site buys. Network means that the ad will be shown on any site within the sales house’s network e.g. any site that 24/7 or Real Media or DoubleClick deal with. A Channel buy costs more as the ad will only be shown on sites that form part of a particular interest channel e.g. entertainment, automotive, sport etc. A Single Site ad is shown only on specified sites and is the most expensive. Illustrative CPM rates might be:

Network: £12
Channel: £20
Single Site: £30

Special sponsorship / homepage / long term tenancy deals would be costed as appropriate. Volume discounts and premium rates are also usual. Sites with perceived high value customers, niches, lots of user data, special areas etc. can charge much higher rates. www.iii.co.uk (personal finance), for example, charges £63 CPM for advertising in a particular area of their site.

NB a third party sales house such as 24/7 will typically only take on sites that generate more that 1 million page impressions per month. They charge around about 30% commission on all ad sales – far higher than offline ad sales e.g. for TV. Some more mature sites do their own in-house sales and ad serving e.g. FT.com to improve their margins and retain control. There have been a number of larger sites that have gone in-house recently. Some have said that you need around 10million page impressions a month for the ROI of bringing advertising sales in house to be worth it.

Pros
It is the standard currency for trading banner ads. It gives a good benchmark to establish site performance over time i.e. a good measure of overall traffic growth trends.

Cons
The figures can be horribly inaccurate. Few sites give out their page impression figures in the first place (either because they consider it sensitive competitive information, or because they don’t quite trust their figures being held up to scrutiny or because they don’t know). Very few sites have a recognised independent audit done to veryify their page impressions: it is a costly and time consuming process. The industry standard currently seems to be New Media Age’s reported figures which are based entirely on what sites tell them.

Use of frames distorts page impression figures: you may see one screen’s worth of a site which delivers one banner ad but is made up of 5 HTML pages sitting in different frames. You will get 5 page impressions logged, one banner ad impression. Non-HTML pages are sometimes not logged e.g. ASP pages (Active Server Pages – created on the fly from data sources) which can mean the page impression figures are lower than the number of pages served. Bear in mind that the number of site page views can be different from the total page view inventory the site has to sell e.g. if there are 2 ads on each page.

Page caching can also cause actual page impression figures to be understated. If a proxy server has stored pages from a site locally then these will be served to the user without a page impression being logged to the original site’s server.

7. Hits

Description
A hit is logged as a request for any file to the web server i.e. not just HTML files but graphics files. Used to be used as the common currency for describing levels of site usage. Often misused now to describe page impressions.

How do you get this info?
Log files…

Pros
Useful as a benchmark of overall site activity over time e.g. if your hits double then it is fair to say your site activity has doubled.

Cons
No real advantage in using hits when you can use page impressions. Hits can easily be artificially increased e.g. a graphic can be cut up into 100 constituent parts = 1 graphic but 100 hits.

8. Browse to buy ratio (or any king of browse to transact metric e.g. browse to subscribe / register / book course etc.)

Description
Percentage of people who visit site that then complete a transaction with the site

How do you get this info?
At its simplest and most obvious, divide the number of transactions (e.g. purchase orders) by the number of user sessions. Some e-commerce software will give you these statistics as part of in-built reporting functions.

Pros
Good benchmark for how successful you are at converting site visitors to buyers. Particularly important if e-commerce is the basis of your business model. Typical browse to buy ratios are around 5%.

Cons
You should bear in mind that many people (quite possible most people currently) will use the web site to research a product and then other channels (stores, phone) to actually buy. Unless you can track the fact that the user came from the web site (e.g. using vouchers, special tel. numbers given on the web site) you risk undervaluing sales that the web site has created i.e. your browse to buy ratio might be higher than you think.