In the past week I have been sent email from two different organizations offering me information about data governance, and both cases seem to indicate apparent minimal dog food self-ingestion.

The first example is actually the last few of a string of emails that I have received over a nine-month period, each of which is addressed to "Jack." In the past month I have gotten six emails about a webinar on data governance. I responded to the sender three times. The first time I asked whether data quality was part of their talk on data governance, perhaps a tongue-in-cheek way of hoping that they'd notice that my email name ('David Loshin") and their salutation name did not match. No response from them. When I got the next one addressed to Jack, I emailed back saying that may name was not Jack. No response. The last email I got from them was responded to with a simpler question: Does anyone actually respond to emails sent to that email address? Apparently not. They don't know Jack ;-).

The second example is perhaps funnier. The salutation on the email I received regarding a new white paper including material on "the inability to use information for strategic business advantage" and recognizing data as an asset to "improve customer experiences" was "{FIRST_NAME}," which is perhaps a little more correct (I do indeed have a first name even if I don't typically use it) although equally indicative of an absence of oversight on the process of producing the information end-product (i.e., the emails).

Last week I had an interesting discussion regarding technical aspects of data cleansing, particularly in the context of acquired data. The challenge posed was that the organization needed to collect data sets from numerous sources with no ability to introduce any types of data controls or dat avalidations. In other words, the data they got was what it was, and if they wanted to use it, they'd have to clean it up themselves.

So the discussion led to talk about tools for cleansing, and I mentioned that most products today provide some means of parsing and standardization as  aprelude to entity resolution, matching, and consolidation. In fact, I will be continuing this discussion at a web seminar next week on Parsing and Standardization, and I hope you can attend!

In the past week, we have had a number of conversations with folks struggling with specific aspects of data integration for master data management. The main issue is that secondary users of what will eventually be master data do not always necessarily bound to abide by the primary users' data definitions. For example, the concept of "customer" means something different to the sales department than it does to those in customer support.

The upshot is that as data element definitions are reinterpreted, the results of sums, counts, and other aggregations start to be skewed. Ultimately, resulting reports are inconsistent, leading to a need for reconcilations, then loss of trust in the master data asset.

One way to address this is a concerted effort to normalize semantics prior to executing the data consolidation. This may shake out semantic inconsistencies and reduce the need for reconciliations.

More importantly, it implies the need for best practices in developing master data models. To that end, I will be presenting a talk on Accelerating MDM Initiatives with Master Data Modeling at a webinar sponsored by Embarcadero on July 28th. Lots of folks have already signed up, and I hope that it will provide an open forum for discussing some critical issues regarding master data modeling.