Dimensions and metrics provide the basis for business relevant information of a Website, out of which result two important consequences. The detailed planning of metrics and dimensions lays the foundation for Web analytics. A good basis for the determination of the used dimensions generates the questions that the Web analytics system

should answer. The answers to these questions will be subsumed for the decision makers in the reporting. After selecting the necessary dimensions, the choice of the appropriate Web analytics tools has to be carried out to provide the desired information.

As the available tools differentiate significantly, the accuracy of the specification is particularly important. The following shows you some criteria that can help to examine the suitability of the tools for the own Web analytics tasks.

Interrelation of metric, dimension and entity

The set-up of your own statistics in the Web controlling does not have to end with the standard metrics "page views", "visits" and "visitors". Most programs in this environment offer evaluations of these metrics according to dimensions such as domain names, Website areas, or navigation hierarchies. By defining own metrics however, mistakes are repeatedly made and it is not rare that metrics, dimensions and entities are simply confounded and jumbled. Thus, the counting of successful registrations represents an own metric. Simple, as this is just a counter that can be displayed over time: yesterday one hundred, today two hundred registrations.

The first and most common problem in this context is that in a “quite simple” implementation of the page only the acknowledgment page "Thank you ..." is counted.

We see, however, on these pages around 15 to 20 percent restorations, reloads, for example through the F5 key or the like, which definitely falsifies the results. Moreover, it is here not about the metric "number of registrations" but the metric "page views" and this makes a difference.

If you want to count the real registrations, you need a close connection to the backend. The same applies for orders and leads or the like. It gets interesting when questions are asked like: "How many purchases are made by which age group in the past month?" Then a dimension "age group" is needed that can be broken down according to the metric “orders”. An entity of this dimension "age group" would be for instance "30-39".

The approach to the creation of metrics and dimensions is, firstly, to formulate the questions from the reporting correctly and then to determine what metrics in which dimensions have to be presented. Only then the necessary technical implementation can get derived from it.

Allocation of dimensions to metrics

Ideally, metrics and dimensions can be allocated freely, but that is not the case with all Web analytics tools. Some tools collect, for example, the visitor characteristics such as age or gender et cetera only in a selection of available metrics, for instance only traffic metrics, although a resolution of these dimensions according to success metrics such as newsletter registrations can provide important information, too.

Correlation of dimensions

Dimensions present characteristics of certain events, but sometimes an attribute alone is not enough to gain the designated information. For a metric "newsletter registration" can, for example, the dimensions of "sex" and "zip area" be defined. But individually, these dimensions still don’t give an answer to the question how many newsletter applications result from female visitors from Manila in the past month. For such a question there has to be established a correlation between the dimensions "sex" and "zip code", whereby only correlations between dimensions that are measured in the same metrics are feasible.

Currently, only high-quality Web analytics tools are able to correlate dimensions, although often only the correlation of two dimensions is allowed. The question of how many newsletter applications have been carried out by female visitors aged 18 to 30 years from Manila in the last month can however only get answered with the correlation of three dimensions, namely sex, zip code and age. The flexibility in the correlations is yet another distinctive feature of Web Analytics tools. In many cases, the correlations are provided and cannot be changed by the customer. But since the choice of the dimensions and correlations has decisive influence on meaningful reporting, you often can’t waive the user defined choice of correlations.


When it comes to reporting, the responsible decision makers and employees should be informed regularly about the latest target achievements and any potential weaknesses that come to attention.

The performance of the reporting has a significant influence on the acceptance among users and thus on the overall success of the Web analytics project. The following aspects help in assessing the reporting functions of a tool:

  • Can dashboards be established in the way that they display the figures individually, for example in relation to the role or function?
  • Can dashboards and reports send automatically scheduled email, for example, PDF files?
  • Is it possible to design data groups and user rights for the access to the reporting?

In addition, the reports should be documented in order to explain "outliers" in the figures.

Key Performance Indicators (KPI)

With the KPI, collected on the basis of dimensions and metrics, the numerous indicators are consolidated to meaningful success factors for the online business.

The following aspects help in assessing the KPI functions of a tool:

  • Can own KPIs get defined?
  • Can objectives (set points) be specified for the KPIs?


Web analytics data form the basis for decision-making regarding possible business optimization measures or idle potentials in the online area respectively. The duties of the performance marketing can be done significantly more efficiently with integrated solutions than with allotted systems.

The following aspects assist in the evaluation of a tool - as for example the integration of SEM activities:

  • Can the performance of search engine marketing activity be measured?
  • What search engines can be administered?

Selection of a suitable Web analytics tool

An ideal “short list” of suppliers leads to at most three different tools which undergo a further comparison.

Experience has shown that a longer list of suppliers and more offers do not necessarily mean a better basis for decision-making. Rather it is crucial to make such accurate information, possibly about the various criteria as for example KPIs, dimensions and metrics that can be integrated into the reporting.

In order to be able to make an appropriate pre-selection of Web analytics providers you should take a look into the shopping guide of Ideal Observer. Based on the criteria listed there a pretty long list can be created, which should then be limited by means of primary criteria on a manageable number of manufacturers - whereas not more than three are really recommendable:

  • Economic position of the provider
  • Underlying license scheme References, active and passive, with preference on solutions that have similar business models as a basis as the site to be measured.
  • Readiness for marketing of the product.
  • Consulting and service competence of the provider

Manufacturers in the still relatively young market "Web Analytics” often tend to an aggressive sales strategy, which sometimes can really lead to "overselling”.

In order to avoid here unpleasant surprises, the tender documents, including Request for Information (RFIs), should describe the requirements as accurately as possible and weigh demand. Otherwise occurs what we name in projects the "100 percent syndrome": All invited suppliers to the bidding meet the requirements one hundred percent.

This effect can be expected if far too banal performance figures are asked in a too general way.

Evaluation of tools

An effective means of assessing the performance of software programs is the evaluation, where for a limited period, in a limited but representative scenario, a number of tools for evaluating the capability are "test driven".

Web analytics deal with the measurement of success criteria on Websites that include in particular those KPIs that show the economic success of Websites, portals or shops. In general, these KPIs aren’t trivial and require adequate expenses regarding the technical integration of the tracking tools. Insofar companies run into two possible problem areas by evaluating such tools:

  • Either the evaluation avoids the major technical effort and instead carries it out on the basis of primitive figures - then the result is not representative and leads again to the "100 percent syndrome".
  • Or the company tries to implement all KPIs with the tools that have to be evaluated, so that the economic framework (ROI) of the project is at risk.

Experience shows that particularly those projects resulted in substandard outcomes, where a large number of manufacturers with trivial requirements have been taken for an evaluation. Project durations of more than one year, with following predictable statistics, were unfortunately the result - apart from an irreproducible economy for such a project handling. Rather, it has to be assumed that the total expenses for the implementation of the evaluation can neither be amortized by a better decision as more economic, nor the selection of a supposedly better provider.