Not too many years ago I was in Paris on business.  As is not uncommon in France, there was a strike in the city while I was there and my meetings were cancelled.  So, what’s a guy to do in the City of Lights during the daytime?  Yep, headed over to the Louvre.

Great place, that.  It was thrilling to see artifacts from 4,000 BC.  And the art!  I hadn’t realized how large some of the masterpieces were – that just made them more grand.  The only disappointing thing about the museum was The Mona Lisa.  She was perched high on the wall behind a thick sheet of protective glass . . . You’ll get a better sense of Leonardo’s masterpiece from photos on the internet, but I digress . . .

Did you know that the Louvre has over 380,000 items in its collection?  Only about 35,000 are displayed at any one time.

Okay, okay . . . I’ll get to the point already.  What does the Louvre’s collection have to do with metadata management?  Plenty.  And it all comes down to curation.

Someone (or in the case of the Louvre, a team) has to be responsible for picking through more than 380,000 artifacts and works of art to display something that can be consumed by museum goers in a reasonable amount of time.  These curators pick the finest, the best, the most interesting, the most beautiful and organizes them to elevate and enlighten every person who walks down the parquet hallways.

When it comes to metadata management, someone (or some dedicated team) has to elevate and educate every user who searches for a data element or who is trying to figure out what product code 3347 really means.  The metadata curator doesn’t pare down the collection, but he or she does keep it organized and up-to-date and accessible and useful to everyone who searches it.

In complex data ecosystems, new data elements are coming in all the time.  Lists of values need constant updating.  Source systems are upgraded and replaced.  New reports and dashboards are built daily.  As business requirements evolve, data mappings need have to keep pace.  Lots of people are involved in these processes – architects, modelers, developers, analysts, testers, report writers, data scientists, and end-users.  But besides the data itself, what’s the one thing that’s constant through that whole chain?  The metadata and the curators who keep it clean and useful.

What makes a good metadata curator?

  • Dedicated time to do the curation – maybe it’s part-time, maybe it’s full-time, but the job needs to be part of someone’s official job description
  • Data knowledge – understanding data structures and formats, data models, SQL, and how the data gets used by everyone in the value chain.  Besides technical skills, needs to have a good grasp of the principles of master data management and data governance.
  • Business knowledge – if you’re in healthcare, you’d better know what SNOMED and medDRA mean.  Banking?  1C2A had better mean something to you.
  • Excellent written communicator – Metadata management and data mapping tools are all about presenting written information to others.  Someone’s got to make sure that it’s formatted, written in a clean, standardized way.
  • Self-motivated, works independently – documentation isn’t usually anyone’s favorite pastime, but someone has to knock this out of the park.
  • Great collaboration and teamwork skills – will be working with and across several teams (often at the same time) to make sure the metadata strategy is effectively carried out by each of these teams.  Needs to help disparate groups come to consensus (try getting bankers to agree on the definition of a “loan”)
  • Anal retentiveness is a good thing to have – incomplete lists of values or missing definitions should drive the curator into orbit.

The metadata curator isn’t just the administrator of the metadata management software.  She’s the librarian that knows how to find everything.  He’s the collector that hunts facts about every system and field he can find.  He’s the editor, constantly clarifying and making things more readable.  She’s the steward who knows that the only thing worse than not having metadata is having incorrect metadata.