WHAT'S NEW ?

Connecting Various Information Systems

Imagine these scenarios:

1. You subscribe to car insurance. In order to process your application, the insurance company generates a contract number which is used as your customer ID in their CRM. Six months later you decide you also need to take out house insurance and you go back to the same company and apply online for house insurance. Like many people you probably have several email addresses and you use a different email account on this new application. A new contract number/customer ID number is generated for you for your home insurance. They now hold two records for you with many of the same details: address, name, credit card details, but two different email addresses. When the time comes to send you their newsletter, which email do they use?

2. You hold an account with a major online retailer where you order books and video games. From time to time, your wife uses your account to sign on and order Dvds, CDs and books in areas that interest her. When prompted, she enters her own email address as a confirmation email address, but it is still under your account. So what happens when they send out emails to suggest products you might be interested in? Do you receive offers for high heel shoes and baby books or for Playstation games and football annuals? Should their system flag up the anomalies in purchasing patterns and confirmation emails and delineate two separate users under the same account?

3. When shopping instore you are offered a loyalty card, which you duly sign up for by giving over your email address. The store is then tracking your instore purchases in their database. At a later date you go onto their ecommerce website where you set up an account. Two things might happen: you might remember you already have a loyalty card and use that number (although you might provide a different email address) or you might forget your existing loyalty card and set-up a new online account. Either way this information will be stored in the ecommerce database. Now when the data is poured into the datamart (central database) they have multiple data for one customer. If they go to run an email campaign, which email address do they use? If the customer has signed up for two accounts with the same email address, can they identify this? How do they remedy this?

Human behaviour

All of these scenarios flag up the fact that human behaviour is inconsistent.

While databases are built on structure, unique identifiers and relationships between data, human behaviour is a bit messier.

We forget. We get mixed up. We do things too quickly. Our circumstances change. People are inconsistent, while databases love consistency.

How do you bridge the gap?

It comes down to planning what data you are going to use for your marketing campaigns and what rules you are going to apply to that data. This is where the experience and knowledge of Cabestan can really help you to plan your marketing campaign effectively. Experience is the teacher of what data to use and what rules to apply to it in order to maximise your customer data.

Let’s look at a few points in relation to data selection. This diagram can help us to understand what is happening:



ETL

Extracting data

To the left of the diagram are some of the contact points where you might be gathering data from your customers. Each of these contact points likely saves information to its own database. Each database may have a different structure or code the data differently.

Your first decision is what data you are going to extract. This will depend on the desired outcomes of your marketing campaign. Right data, right time will be determined by your marketing scenario. Rules are then set up in the marketing campaign management tool as to what data is extracted from which database.

Transforming data

The next step is transforming the data. This step looks at what rules will be applied to the data so that it becomes useful for your scenario. As the data may not be stored in the same way from one database to the next, you may need to look at transforming it so that you are comparing apples with apples rather than apples with oranges.

At this point, you may also want to look at deduplication of the data – a process which helps to eliminate duplicate data in your database and solve many of the problems highlighted in our opening scenarios.

This is achieved by running queries for groups of records which share similar traits and examining them further to see if they are duplicates. For example you could run a search comparing last names and addresses looking for duplications. If that came back with too many records you could add telephone number to the mix as a way of refining the results to be able to find and eliminate duplicate records. You need to test for the right combination of comparisons in order to get the best results.

This process can also flag up anomalies in records such as multiple email addresses. A rule can then be set as to which email address will be used. This will depend on your marketing scenario. If you were looking to push online sales, then you might favour emails collected through your ecommerce site, whereas if you wanted to push store sales you might favour emails collected instore.

Loading data

Converted data is now entered into your marketing datamart ready for your marketing campaign.

With automation, and the right skills to set it up, this process can happen quite quickly and before you know it your raw customer data is ready to drive your next marketing campaign while dealing with some of the inevitable pitfalls mentioned in our scenario cases.