Friday, January 3, 2014

Data-Driven Virtuous Cycles

To paraphrase the Six Sigma religious tenet, "you can only manage what you can measure." While I think that this is somewhat over-simplified, there is certainly truth to it.

Specifically, I run into issues which can only be resolved with a solid metered context. Often, while debugging these issues, I build "scaffolding" which then lives on as the metering to drive a continuous monitoring process.

Some issues have many factors, related in murky and maddening ways and the only way to untangle the knot is to measure, find and fix something and then return to step one until your measure tells you that you are done.

Current Example
We are in the process of restoring functionality lost when a highly tuned system was replacing by something manifest worse--but cooler. One of the data elements lost was who collected the specimen. This turns out to be critical for many management functions.

The first reaction to our bug report was "nonsense! the new system is fabulous!".

The second reaction was "ok, looking at your data, we see that this is happening, but we have come up with a labor-intensive workaround and we will simply command people to follow the new procedure."

The third reaction was "ok, we see that some areas are not complying but we are going to scold them--and stop recording this data, because we don't need it anymore."

Needless to say, we are still collecting the data and still helping them police their spotty compliance. Someday, the meters will tell us that all is well and we can go back to relying on this data for our high value reports.

The Bad Old Days
This situation is sadly similar to our work with scanned requisition forms. When we deployed our draw station solution, we became part of the scanned req infrastructure. As the newest member of the team, we were immediately blamed for any and all missing reqs. In self-defence, I created an audit to trace scanned req activity, comparing expected with actual. We immediately made a number of interesting discoveries:
  1. I had some bugs, which I fixed and verfied as fixed
  2. Some users were not really on board with scanned reqs so we started to nag them
  3. Some of the orders for which we were blamed did not come through the draw station; the Front Bench decided to use our software to ensure compliance
  4. Some of the scanners were in need of service
  5. The placement of the bar codes on the page matters more than one would hope
With feedback and monitoring, the situation has improved dramatically and our req watchdog technology is actually still in service even as the LIS and draw station solution for which it was created are out of service and about to be retired, respectively.

Tube Tracking
I think that our tube tracking experience can also be seen as measurement leading to clarity and control, so I am including it.

Measure, management, repeat. Even when all is well, don't stop auditing and reviewing.

No comments:

Post a Comment