Key Performance Indicators (KPIs)

NOTE: This is an ongoing original fiction story that I’m currently writing. I started writing this fictional story back on October 2, 2020 and contribute ~1,000 words to it every day on this blog. I didn’t outline the story at all going into it but it’s slowly evolved into a tale about a data scientist in his mid-thirties from America who finds himself summoned to China where’s he’s been offered a job to work for the Chinese Communist Party on a project monitoring the Uyghurs in the Chinese “autonomous region” of Xinjiang. In China, the story’s protagonist, Dexter Fletcher, meets other professionals who’ve also been brought in from abroad to help consult on the project. My story takes place several decades in the future and explores human rights, privacy in an age of ever-increasing state-surveillance, and differences between competing dichotomies: democracy vs communism, eastern vs western political philosophies, and individual liberties vs collective security. If this sounds interesting and you’d like to read more, my fiction story starts here.

Chapter Eight – Passage Six

KPIs –or Key Performance Indicators– are the critical benchmarks that any data science project is based upon.  For example, back in the US on the other projects that I’d slaved away on in the past, KPIs were often mundane metrics that you’d pretty much expect from any pedestrian, vanilla project:  The number of new people who enrolled for healthcare during November, the amount of advertising revenue that a marketing campaign was currently generating, etc.

In our case, since project was a bit more exotic, the KPIs of our assignment were likewise more exotic as well.  We measured our success in Xinjiang among two primary metrics:

The first was a scorecard gauge of the number of crimes committed in Urumqi on a given particular day.  Crimes obvious comes in all flavors of the rainbow –from petty theft to arson to murder– but for our overhead reporting purposes, we had a single summary statistic that aggregated all crime numbers.

By the way, I should take a slight detour to mention here:  In data science, the devil is entirely in the details.  There’s a famous saying in our profession:  “All models are wrong.  But some are useful.”

At the score of data science is the desire to make sense of reality around us with numbers– to somehow quantify the ineffable.  In a case like looking at the crime statistics in Urumqi, we needed a single number to summarize how are policies were performing in the capital.  But if we instituted a policy that decreased petty theft but increased murders in the city, was that a win?  All crimes are not so obviously we then need to weight these metrics somehow.  But how, and who, determines that?  Does every murder equal five incidents of petty theft?  Ten incidents?  Etc.

As you can see, the entire project quickly turns into a scenario modeling and analysis exercise.  For example, we’d devised two models to measure crime differently.  Crime, in China, is broadly categorized under three classes:  Trivial (Class 1), Moderate (Class 2), and Severe (Class 3).  For example: Trivial would be your petty theft or drunken pub brawl (where no one was injured); Moderate would be the vandalization or destruction of property; Severe would be murder or inciting subversion of state power.  (Notably, in China, assembling in groups larger than fifty people required a local municipal permit.  For instance, a wedding with over fifty guests?  You’d need a permit for that.  And violation of this mandate would result in a Class 3 violation of Chinese law which carried a hefty fine and, depending on the kind of meeting, imprisonment or even death by execution.)

Our first model had a 3x multiplier for Class 2 crimes and a 10x multiplier for Class 3 crimes.

Our second model featured a 6x multiplier for Class 2 and 15x for Class 3.

But our models were consistently failing to manifest real-world results that we expected.  Week after week, the scorecards on our dashboards remained unchanged (or went the wrong direction!); despite the fact that our analyses and models had predicted certain outcomes.  But for some reason, in the real world, we were not achieving our KPIs.

After Alan explained the way that we’d set up our models to Jack, Jack had just thrown back his head and laughed.

“Making Urumqi a totalitarian police state, despite whatever you may have been told, is never going to work.  Occupation simply breeds hate and resentment which’ll fester.  Maybe quietly at first, but make no mistake.  It will most certainly boil over.”

“So what do you suggest?” Kristen asks, irritated.  I also felt my own collar growing hot.  Who was this lazy bum to lecture us on our efforts?  What did he know about suppressing minority populations in communist regimes?

“My thought,” Jack says, “is you loosen all of the restrictions.  Withdraw.  Give it a year or two.  Hell, give it maybe six months.

Alan stares.  “What?”

“All of the electricity and civil services in the region are entirely reliant on Beijing,” Shu says, “without a Chinese presence, the entire area will degenerate into complete anarchy in a matter of weeks.  Supply chains, crops, clean running water…”

Jack waves his hand.  “So what?  The Uyghurs want freedom?  I say, give it to them.”  He turns to me, “you westerners have a saying, do you not?  The grass is always greener on the other side?”  Jack laughs.  “There is no grass on the other side!  Or if there is, it’s all yellow, dying, and dead!”

I turn to Alan, “Just out of curiosity, what would have happened if Beijing withdrew from the region?  This is nothing we’ve never modeled, right?”

Alan frowns.  Of all of the hundreds of scenarios that we’d entertained and tried over the weeks, simply giving up and going home had definitely not been anything that anyone had thought of.  His forehead creases in that way which always happens when he’s consternated.

“Well,” Alan says slowly, Kazakhstan would most definitely see a withdrawal of that magnitude.  They’d most definitely be shocked.  It’s been over a century of contesting that geographic region.  To suddenly pick up and just go home…”

Coleman interrupts.

“Guys, wait up.  You’ve all just spent weeks telling me how Xi and China is the most honor-bound society on the planet.  Even if this plan somehow yielded results, which is still dubious to me, what on earth makes you think that Beijing will go along with this?  Wouldn’t this be an ultimate sign of great shame and surrender?”

“I’ve got it.” Deepak says suddenly and we all turn to him.  “Coleman’s right, of course.  Beijing will never just withdraw from the region voluntarily.  But if we manufactured pretext… if we somehow, someway provided a reason to withdraw…”

“…you mean something like a natural disaster?” says Shu.  She taps her fingers against her lips.  “Something like–“

“–something like a man-made disaster,” Kristen finishes the thought.  I realized what Deepak was getting at a split second before Kristen did but it makes sense.  In the most awful, frightening way possible, it makes sense. 

“You want to manufacture a kind of biological crisis,” I say aloud.  “In one fell swoop, it would solve all of the problems.  Some of the old guard, which has been the most resistant to Chinese control, would be the most vulnerable.  And the new generation, the most politically active, have the softest hands the world has ever seen.  China’s weened them for years now to use smartphones and computers– this a is a generation that couldn’t milk a cow or farm agriculture if their lives depended on it.”

“By withdrawing Chinese support for maybe a year after they were to accidentally receive some kind of plague would send the entire region into chaos,” says Alan, “hundreds of thousands would likely die.  Maybe more depending on the potency of the biological agent.”

Jack nabs the final shrimp dumpling from the dim sum bowl.  The old man honestly looks the most alive I’ve seen him since we’d arrived on our trip yesterday.  He chews slowly and thoughtfully, and then swallows.

“Maybe,” Jack says.  “But you’ve gotta admit– it really could be the answer to all of your problems.  It really could.”

Leave a Reply

Your email address will not be published. Required fields are marked *