donderdag 27 februari 2014

Double last names for newborns naturally increasing Data Quality?

The Belgian government is currently discussing around a new controversial law giving all newborns a double last name/surname (name of both parents in case both are known) by default.  
This is also referred to as double-barrelled names.

Although couples can still choose to pick one of both, this could make names all in all more unique over time.

In case 2 (future) parents (both having a double last name) have a child, that child would have 10 optional last name-combinations (vs. two right now (father's or mother's last name)).

Stating that there's a 5 times less likelihood to get duplicates is overrated, but it will surely increase uniqueness to a certain extend.  Some people will stick to one last name while others will only provide us with one of both.

Filtering out double-barrelled names with some simple pattern detection will allow us to detect relatives which is usefull in e.g.: telecom to recognize family-bounds etc.




woensdag 5 februari 2014

Do different types of businesses benefit from different types of MDM?

Intro & Question:

Besides the different implementation styles of MDM, there are also a number of types of MDM.

We've seen that landscape complexity and MDM/MDG maturity can influence the implementation styles, but what about the type of MDM that's chosen - is that linked to the type of business you're in?

Typically we distinguish 3 types of MDM: Analytical, Operational and Collaborative, used for different purposes (see wiki below).

Do different types of businesses tend to choose for different types of MDM:

  • Will more service oriented companies rather choose for analytical MDM and more 'manufacturing' oriented companies rather choose for operational MDM?
  • Or are all companies leaving the idea of analytical MDM and going with operational MDM, connecting it to (large) data warehouse (appliances) & running analytics from there?


Wiki:

Implementation styles of MDM:
  • Consolidation
  • Registry
  • Co-existance
  • Centralized (transactional hub)

Types of MDM
  • Analytical MDM solutions contain data in a more BI-fit & aggregated manner, making it easier to use for BI-applications & analytical purposes.  This type of MDM can often have BI-specific hierarchies and will usually not provide the lowest level of granularity (compared to operational MDM).
  • Operational MDM solutions contain data on the lowest level of granularity required for the operational consumers to perform their everyday transactions and business processes.  The data is not necessarily aggregated for analytics (perhaps a little for operational reporting).
  • Collaborative MDM deals with the creation of master data in support of a business process.  Nevertheless, since analytical and operational mdm are most used & known, we will concentrate on those in this article.

Use:
  • Using analytical MDM for operational purposes & in operational consumers can often cause issues because of the level of granularity & level of aggregation.
  • On the other hand, using an operational MDM source for analytical purposes, might not provide the required performance, but will do the trick (Performance issues can easily be overcome these days by means of (E)DW's, loaded through a good ETL).