There is No Governance without Data Governance

I try and listen to most of the White House and Covid 19 Task Force Press Briefings. I also listen too President Trump’s interviews. The last interview with Axios’ Jonathan Swift re-enforced a recurrent theme that pops up in most of the press briefings and interviews. There is no agreement on what really matters when determining how well the country or the government is doing in responding to the pandemic. This is at the heart of sound of data governance regardless if it one is guiding a company or a country. In this age of information, data governance is all.

The Axios interview illustrated the importance of governance on multiple occasions as the President and Mr. Swift tried to use different measures and different data sets to make their points. At the end, much of the interview disintegrated around the fact that neither took the time to agree on what was important. Yet, understanding what is important is the first step in establishing sound decision making and sound data governance. Not having that common understanding leads to conflicted decision making.

For a company, understanding what is important might be a relatively easy thing to identify. But when faced with a pandemic, determining what is important isn’t quite as clear. Is it more important to slow the spread to some minimal level with a larger impact to the economy in the short term, or should the economy and the spread of the virus be weighted in some way? There is a similar trade off between having students attend a physical school versus risking community spread. Until leadership decides what is important and ranks the different classes of importance you can’t really collect data efficiently or decide what to analyze. Everything is sub-optimized.

Presupposing that there was agreement that the most important objective was to slow the spread of the virus, the next question is how do we measure progress towards this goal. This is not a simple answer. Is it the number of new cases per day? Is the rate of change of new cases? Is it a proportion of new cases to total population? Is it the number of total tests versus the rate of confirmed infections? Is it US numbers benchmarked against other countries? Without clear goals there is no way to measure success or failure of tactics and strategy used to achieve goals. The interview stalled over this debate as the two participants threw graphs at each other and debated the meaning for lack of clear measures.

The next foundation of sound data governance is the management of data. The Axios interview only hinted at this issue when the two participants compared US statistics to those of China, the EU, and South Korea. Both participants agreed that some countries’ data was suspect. For me, US data is suspect; suspect and incomplete. All one has to do to understand this is look at raw case data reported by counties and states. Some states have clumpy data where they aggregate and report periodically. There are days when total cases down. There are differences in the fidelity of tests, the type of tests, and times between when the test was administered and when the results were available. To a statistician or an analyst, drawing inferences would always be questionable especially when looking at specific points in time. That is why most of the reporting done by the CDC and others uses rolling averages. Yet precision is important.

The final governance area that stood out was the lack of quality analytics and the failure to establish true causal relationships. The President repeatedly claims, as he did in the interview, that his decision to halt travel from China saved millions of lives. On its face, the claim makes sense. Fewer people coming from Wuhan means less people to spread the disease. But is that really true? 

The first cases of the Virus were detected in early January. On January 23rd, the Chinese government closed down the city of Wuhan. But it was already too late. On Jan 31st, President Trump restricted (not banned) travel to the United States from China. The first case in the United States had already been reported. Three weeks later, the US became the world leader in new cases. The virus had entered the US from multiple vectors. 

Designing decision support systems requires real thought. In measuring our performance against the virus, we want to know more about causal relationships. What was the effectiveness of complete shutdowns? Why when the nation was on shut down except for essential works, did the virus still spread? What about understanding the effectiveness of wearing masks or gloves or face shields? How about the effectiveness of new treatments? This is where data governance, data science, and data analytics all factor together to determine the effectiveness of both decision support and policy effectiveness.

When I reviewed this short (69 second) dialog, I just cringed.  

Jonathan Swan: (13:46) - Oh, you’re doing death as a proportion of cases. I’m talking about death as a proportion of population. That’s where the U.S. is really bad, much worse than South Korea, Germany, et cetera.

President Donald J. Trump: (13:55) - You can’t do that.

Jonathan Swan: (13:56) - Why can’t I do that?

President Donald J. Trump: (13:58) - You have to go by where… look. Here is the United States. You have to go by the cases. The cases are there.

Jonathan Swan: (14:04) - Why not as a proportion of population?

President Donald J. Trump: (14:07) - What it says is, when you have somebody where there’s a case-

Jonathan Swan: (14:11) - Oh, okay.

President Donald J. Trump: (14:11) - The people that live from those cases.

Jonathan Swan: (14:14) - Oh. It’s surely a relevant statistic to say, if the U.S. has X population and X percentage of death of that population versus South Korea-

President Donald J. Trump: (14:22) - No. Because you have to go by the cases.

Jonathan Swan: (14:22) - Well, look at South Korea, for example. 51 million population, 300 deaths. It’s like, it’s crazy compared to-

President Donald J. Trump: (14:28) - You don’t know that.

Jonathan Swan: (14:29) - I do.

President Donald J. Trump: (14:30) - You don’t know that.

Jonathan Swan: (14:31) - You think they’re faking their statistics, South Korea? An advanced country?

President Donald J. Trump: (14:33) - I won’t get into that because I have a very good relationship with the country.

Jonathan Swan: (14:36) - Yeah.

President Donald J. Trump: (14:37) - But you don’t know that. And they have spikes. Look, here’s one of-

Without Data Governance there can’t be good governance.