OLAP data cubes are the only way to go for any organisation that takes web analytics seriously. Even though I am a big fan of Google Analytics, retrieving non flattened, customer specific data from it is impossible. And even though, in theory Omniture should do the job for you, in practice (in my experience and that of my colleagues) it fails do deliver data accurately and reliably most of the time. The good news is, that if you can afford to think about omniture, you can afford to get data cubes!
What Are Data Cubes
OLAP (online analytical processing) is simply a spreadsheet which is refresh-able like a pivot table (whilst accessing real time data from your reporting server). In the back-end what is happening is an SQL query looks up and joins dimensions and metrics from your reporting database and sends back the results to your excel spread sheet. A cube has the ability to access two dimensions such as product and city and display them with multiple metrics such as conversion rate, revenue etc. The best part of having the reporting infrastructure that runs an analytics data cube is that you can spin out as many cubes as you want with different dimensions and metrics really easily. This means, that if you have multiple marketing channel owners who need to view information on an ongoing basis that is slightly different to a standard report – you can simply spin out a cube for them. Similarly, cubes can be spun out for share holders of different seniority – for example the marketing director may want to view the performance of all marketing channels over time, whilst the product manager may want to view SKU performance by customer segment. All of these are easily solved for within a proper reporting framework.
How Marketing Data Cubes Can Be Implemented
Every ad, keyword, placement, creative, channel, targeting option and marketing partner can be tagged with a unique URL. This is standard digital marketing practice. That way, you build out a hierarchy of channels, partners and placements and get to view and optimise them based on performance. With this insight, you can stop wasteful spend easily, and, by observing data trends over time generate accurate forecasts into the future.
At a practical level this looks something like:
You have your URL: www.coltrane.com.au
You have a coding convention for tagging the URL: www.coltrane.com.au?channel:placement:creative – the question mark can be used to signal to your webserver to capture all data after it and store it in relation to the activity of the user on the site.
For example, say you were advertising on hotmail with a creative called ‘monkey’ – your URL may look like www.coltrane.com.au?ninemsn:hotmail:monkey – then all purchases, user behavior such as bounce rates, page debth and interactions on the site can be recorded and attributed back to that marketing creative, placement and channel.
Own Your Data
The biggest difference between hosting your own analytics and letting someone else do it for you is that you own it, have private access to it, and can use it how and when you want. This cannot be said wholeheartedly for any 3rd party analytics platform.