All the technical detail, expertise and advice


Celebrus Implementation – fast but not furious

Published: Monday, 29 April 2013 10:00
Big Data Analytics

Over the years a frequent topic when talking to prospects is "how long does it take to deploy Celebrus?" This is not an unusual question, mainly because most of the online data collection products currently on the market rely on tagging strategies to gather the more valuable data from a website.

Typically this deployment can take weeks, months, or even years to complete, and frequently means that an organisation has to compromise on its data quality in the interests of time and money. In reality, the tagging process probably never actually stops as most websites will change over time, so time and, therefore, money is always required to keep these systems collecting the right data.

In comparison, the total time taken to fully deploy a Celebrus system is relatively very short, typically taking between five and fifteen man days to complete, depending on the type of system being deployed.

To prove the point we recently performed our own use case scenario to reconfirm exactly how long it takes to install, deploy and configure a Celebrus Core Collection Server (CCS) and Audience Behavioural Intelligence (ABI) system, to produce a customer data mart containing details of visitors and their actions performed over many visits.

It actually took no more than five days to complete from start to finish. Wishful thinking? Let me explain how we did it.

We were ready to begin installing Celebrus software from the start as the hardware was pre-prepared. Installation of the software is not an onerous exercise, especially if you have the relevant information to hand; such as database access details, SSL certificates and DNS address details. It is having to look for those details that can cause un-necessary delays.

For both the CCS and ABI, no more than two hours was required to install and test each component which, in our case, was one Collection server instance and one ABI server based on our expected levels of web traffic.

If you are familiar with traditional tag-based technologies, then you might be very surprised to hear that this time also included the time taken to deploy Celebrus’ unique insert on the target website. The Celebrus insert is identical for each and every page so can be added to the footer template of your website (or websites), or injected at the network layer, meaning deployment takes just a few minutes.

The next step was configuring the collection system to gather other useful data elements such as individual product prices, basket values and purchased basket values. For us to gather this kind of data we simply configured the collection system with no additional tagging required. In our case it took just a few hours to identify the required elements, configure the collection system and test that the correct data was being collected as expected. Job done!

The final phase in the deployment was configuring the Analytics server (our data processing system) to convert the raw data that is being collected into business events and customer profiles. This was the most time consuming part of the exercise which took around four days in total to complete.

This phase was completed in two distinct stages. The first stage required assigning business meaning to the raw actions and events collected from the website. For example, we defined what a product view looks like, how to identify when someone views their basket, adds or removes items from it, or successfully completes a purchase.

In the second stage, these business level events become the basis for defining the significant behaviours that we want to record for each and every online customer and defining how we identify who these customers are. In our case we used details gathered from data entered into a user registration/login form, but we could have used information extracted from site cookies or content embedded into the page by the Content Management system if needed.

And that was it - from nothing to streaming, highly valuable long term customer data within five days. The best part is that this is allof the work that will be required to deploy the system, because from this point onwards the system begins the process of building a customer database totally automatically. This includes maintaining your customers’ identities over many different sessions and devices to build a true single customer view, calculating the RFM scores and overall monetary values for each and every customer, and monitoring ongoing customer defection status, amongst many other insights that can be developed based on the Celebrus data.