Data Quality & User Overrides: Shut Up and Take My Money!


Stephen Pace

Last Tuesday a data quality vendor put out a blog post “Companies need to be educated on data quality“ that essentially said that companies need to be educated in data management.   While one of the surveys mentioned focused on financial services organizations, I think the same can be said of many industries, and you certainly don’t need be a data quality professional to see data quality problems everywhere in daily life.  What’s more, I think these types of everyday stories are the things that resonate the most with business users when describing the problem you are trying to solve for them.   Some of these issues are merely annoying, but often these data quality issues translate into real money lost or at the very least customer goodwill expended.

Over the years I’ve seen more of these than I can count.  One CPG company we worked with had changed operational systems and in the process, mis-mapped some reference data.  This caused an ingredient misorder and some rather unpleasant material to show up in one of their manufacturing facilities.  Just in time manufacturing is great—until you have a stock out and have to shut down the line until the real ingredient arrives.  The sad part was the people that placed the incorrect order realized it was strange, but ordered it anyway because they trusted the system.  Good master data systems allow participants to raise issues when they see something may be amiss and routes those requests to the people that can either can either resolve the issue or provide reassurance that the data is indeed correct.

This brings me to my latest brush with bad data.  I generally prefer to order things online—too many times I’ve gone to a store that is supposed to have a hard to find item and either inventory is wrong and they are out or the item is a slightly different model than what I wanted.  At the very least, I probably would have had to spend way more time than I’d like scouring the aisles until I finally found it. I’d much prefer to use my web browser to search for exactly the item I want, read reviews, compare prices, and ultimately order for delivery.  It is extremely annoying, then, when you get to the point of pulling out your credit card and the system doesn’t let you proceed.

City & Zip Code validation error

 

I’ve had something similar happen before.  I used to live in Sugar Land, TX, and my ZIP code changed from 77478 to 77498.  Despite the Post Office warning vendors years in advance, and there being a generous grace period, many processes broke.  For instance, for security, many online ordering systems require the ZIP code of the ordering credit card to match the shipping address.  However, if your credit card company updates their address for you before the store shipping address validation is updated with the new ZIP code, you’ll get a mismatch.  Either you can make the shipping address valid, or the credit card address valid, but not both.  In some cases, even a call to the store’s order line wasn’t enough—I’d wait on hold, finally get a person, only to be told “sorry, the system doesn’t allow me to enter in a new ZIP code, I’ll have to escalate your problem.”  I’m all for systems attempting to improve data quality by preventing bad data entry—but if your customer is telling you, look, I know my own ZIP code, you can go to USPS.GOV and look it up if you don’t believe me, there should be some mechanisms to override.  At the very least, for data problems that inhibit immediate purchase, the workflow and change request process should have a strict service level agreement to get resolution in hours if not minutes.  Otherwise, a sale could be lost.

In the case of a certain big-box store, though, there was no ZIP code change.  As far as I know, my street has had the same ZIP code for over a decade.  As I could proceed no further, I used the Contact Us button to report the problem.  The next day, I received a note saying that they escalated the problem (but no word on how long it might take to resolve the issue) and recommended that I visit their nearest store.  I knew that store didn’t have this item in stock, and didn’t relish driving down to the store and trying to order it.  I then thought I’d see if they might respond faster to social media:

Big-box tweet

 

Okay, I probably could have phrased that better since I wanted a resolution, but alas no response of any kind.  Say what you will able Comcast, they respond IMMEDIATELY to their  @ComcastCares twitter account!  Over the next few days, I started to think about what might cause the issue.  Was my ZIP code now associated with a different city? I checked USPS but everything looked fine.

I’d heard nightmares from people where the official data providers had accidentally screwed up their address.  They would attempt to update their address for all of their bills only to find out the next month, they had all automagically reverted to the incorrect address because the system trusted the purchased data over that which got phoned in.  MDM systems need the ability to have any number of status fields such that automatic update processes can be told to ignore such problematic addresses once they have been pointed out.  Good MDM systems can also notify stewards if something important like a ZIP code city changes or becomes unassociated altogether.

A week went by.  Like any shopper, during that time I checked other providers (fortunately for this particular store, their competitor doesn’t carry this particular item).  I heard nothing from them, but I eventually tried again.  This time, a different result!  The address data still had an issue, but now I had the opportunity to override the warning and move further:

Address validation error

Success, item ordered!

I still don’t know where the problem lies, but hopefully the store is still looking into it.  And, if they are reading this, I’m happy to talk more seriously about your current MDM system(s) and the processes you have around them.  I believe systems should:

  1. Expect change—few things stay the same, especially ZIP and area codes.
  2. Enforce service levels—if something interferes with ordering, it should have an extremely high priority and workflow should route resolution right to the person or group who can fix it.
  3. Permit overrides—no data is perfect, even third-party purchased data.  Allow people to raise issues and change requests when problems arise and keep an audit trail of what changes were made, if any.

This type of problem certainly isn’t unique.  Many, or maybe most, other retailers have similar issues with systems that can’t simply react quickly to change.   For those who got to the end of this rant, what are some of your favorite data quality issues you’ve run across?  How long did they take to resolve?

Tags: , ,

One Response to “Data Quality & User Overrides: Shut Up and Take My Money!”

  1. Oz January 17, 2013 at 6:55 pm #

    Stephen GREAT post. I love it “shut up and take my money!”

    My favorite data quality issue is from the perspective of someone who scrubs data.
    Marketers who set up user interfaces are a problem waiting to happen.

    I scrubbed several hundred rows of data because a marketer set up a form with tons of required fields. All he really needed to know was that people who signed up for a free newsletter were over 21 years old because it involved alcohol. Yet, their addresses were required, income range information, phone number, occupation category, etc. ALL of it required.

    Of course, the result was a mess. Lots of phone numbers 111-111-1111 or 123-123-1233

    This says nothing about the people who either
    1. Abandoned the form because it was too snoopy
    2. Abandoned the form because they got tired of trying to submit it only to have it come back saying that they omitted certain required data.

    Ultimately, what did this marketer have to analyze?
    - Could we trust that 30% of members are 30 to 45?
    - What about the handful of people who are over 100 years of age? (At least they’re over 100 based on the data in the spreadsheet.)
    - Hmmm … if we call 111-111-1111 could we reach all 20 of the people who said that’s their number?

    That was the first time I saw the other side of those forms. Usually, I’m putting in crap data myself because someone is asking for too much. That experience showed me how worthless the data was because of the required fields.

    It would have been really interesting to see how many people input data any way had it been optional.

Leave a Reply