Summary

COVID-19 burst upon the world in March 2020. At that time, in a series of blogs, we contended COVID-19 was an operations management challenge to avoid data-driven disasters. Over the past six months, significant progress has been made in understanding COVID-19 and reducing its health and economic impact. We contend a good part of this success was the adoption of sound operations management methods – albeit not labeled that. However, substantially different views exist on what should have been done and the best course forward. These differences relate to the following issues:

  1. Gaps in the data
  2. Same data and models, but different interpretations
  3. The interplay between policies and beliefs – upsetting the social order
  4. Intermittent or sporadic events

We demonstrate these issues cannot be solved simply with the application of analytical methods given the current data gaps and the inherent nature of the policy. These same challenges, albeit in a different environment, can impede the success of any investment in new technology to improve supply chain management decision making. A successful implementation requires strong leadership that can navigate the political terrain to build consensus supported by agents of change who harness community intelligence.

Introduction

COVID-19 burst upon the world in March 2020. At that time in a series of blogs, I made the argument that COVID-19 should be seen as an operations management challenge (where public health, epidemiology, and medicine were just components of the solution) and simply following the limited data and science available could generate data-driven disasters. Additionally, the application of operations management would be critical to identify and guide the collection of the required key data, and the scientific work needed to move forward. That is the purpose of a model is insight, whether an organization is tackling COVID-19 or such supply chain issues as estimating demand or matching assets with demand.

Six months later, the nation has come a long way in what it knows about COVID-19 and how it approaches it from limiting the spread to treating those that get COVID.19. Death and hospitalizations are down dramatically in NYC, the epicenter of COVID-19 in March and April. In Arizona, a hotspot that emerged in June, hospitalizations and the number of positive cases has dropped dramatically by September. Physicians know considerably more on how to treat hospitalized patients with COVID-19 and new methods are emerging. Overall the application of operations management principles has put in place structures to keep the risk down from flying to protecting the elderly. There is a plethora of information published daily on COVID-19 and it is certainly a hot topic with respect to the upcoming November elections. The purpose of this blog is not to rehash this information, but to identify open issues in handling the current challenge of tackling COVID-19 and relate this to the challenge of managing supply chains. The issues are:

  1. Gaps in the data
  2. Same data and models, but different interpretations
  3. The interplay between policies and beliefs – upsetting the social order
  4. Intermittent or sporadic events

Gaps in the Data

Although substantial strides have been made in collecting more and better data for COVID-19, and technology is being harnessed to support tracing, substantial gaps remain. Some are:

  1. The actual COVID-19 test result. Currently, each COVID-19 test result is converted to a binary value (negative or positive); this summarization results in the loss of critical information for analysis.
  2. How Seriously Ill? The nation captures the number of positive and negative COVID-19 results, hospitalizations, and deaths. For those hospitalized there is little data collected and stored about how seriously ill a person is. With the ability to better treat COVID-19 patients to avoid death and the shift to younger people being more likely to be COVID-19 positive, deaths have decreased. This can lead to the assumption that the health impact of COVID-19 is disappearing. The lack of information on “how seriously ill” fails to provide a true picture of the impact of COVID-19. At age 66.5 (which puts me in the high-risk group), if I get COVID-19 I am certainly happy if I do not die, but a one year recovering and permanently damaged lungs are still a serious impact.
  3. Testing for COVID-19. The ability to test is on the increase (from a fully functioning test process to general availability), there remains the challenge of random testing and substantial differences of opinion on when testing should occur.
  4. Lack of a single source of all data. The nation still lacks a single source for all critical COVID-19 data. Additionally, there are now turf wars for control.

Same Data and Models; Different Interpretations

An implicit assumption in SCM is once we have a common set of data (more general description of the situation), then analysis (analytics) can drive to an interpretation or assessment of the situation that is commonly held. In my experience, this is not true, and we see this firsthand in COVID-19.

  1. The value of masks. There are considerable differences of opinion on the value of masks that range from differences between the head of the CDC and the president to information and heated discussions in small gatherings where each side points to their experts’ (often with Ph.D. or MD) pros and cons. The debate continues on a regular basis even for motorcycle rallies.
  2. How many deaths were caused by COVID-19? At the end of August, the CDC as part of this regular process of improving data quality identified most people who died from COVID-19 had other health issues (comorbidity). This then generated two different data interpretations.
    1. The vast majorities of deaths that had been associated with COVID-19 were not from COVID-19, but from other ailments, and the nation was intentionally misled about the danger from COVID-19.
    2. The CDC did a review of the COVID-19 deaths and determined that in most COVID-19 deaths the person had one or more underlying health conditions. For example, my mom has limited capacity in one lung from what appears to be a fungal infection. Far from life-threatening and does not keep her from having a fruitful life. If she were to get COVID-19, her risk of death is higher, and if she died, then CDC would note this underlying health condition. But this condition was not the cause of death. That said, attributing a cause of death and the path to death is often challenging as a person ages.
  3. Was the model wrong? In March the “Imperial College Model” estimate that if no adaptions in patterns of interaction (mitigation) were made, the U.S could see 2.2 million deaths. Currently, the number of deaths is under 200,000 and the model has come under intense criticism and some have called it fake to perpetuate the crisis and justify lockdowns. There are views that
    1. The model was a high-level model designed to provide insight where the impact of different responses could be tested. The 2.2 million estimated assumed of no or limited mitigation actions taken. For example, no changes in assisted living procedures (which did happen). The fact that ONE of its projections of 2.2 million deaths was wrong is a function that numerous mitigations efforts were instituted.
    2. The model was structurally incorrect from the beginning and the flaws were ignored.

The interplay between policies and beliefs and intermittent events

Over the last six months, it is clear there is a substantial interplay between the policies put in place if the policies are followed or ignored, and a group’s beliefs or assumptions. For example

  1. The perception of risk from COVID-19
  2. Perception of risk from government restrictions on what were previously day to day activities
  3. Perception of the relationship between restrictions put in place or requested and reduction in risk and economic impact
  4. Differences in self-interest
  5. Differences in what one believes is the best path

Intermittent or Sporadic Events

The difference in opinion is extenuated since COVID-19 “negative outcomes” are intermittent or sporadic and lagged. Most people have not had COVID-19, only a limited number get seriously ill, and fewer die. Seriously ill is not generally reported in any detail. Where COVID-19 spreads rapidly not everyone gets COVID-19. When COVID-19 positive numbers increase, there is a lag time with some random variation between hospitalizations and deaths.  Additionally, as science and medicine improve serious illness and deaths go down from these insights. Success in managing these events requires a shift in orientation to risk management from optimal efficiency. This requires keeping place contingency assets that may never be used or used rarely. For example, carrying “extra ICU beds”.

What is the lesson for Supply Chain Management (SCM)?

The challenges faced in creating and implementing a unified course of action to minimize the impact of COVID-19 are the same faced in an effort to improve an organization’s SCM processes and technologies defining SCM as any aspect of planning and scheduling.

Gaps. There are always gaps in data, some are easy to close, others require work, and some will not be closed near term. Therefore, patches are put in place. Often a difficult challenge is assessing work in progress in production. What fraction of the production start will be “good” and when will it be available?  A second is estimating “jump” changes in demand.

Same Data and Models. The folklore in SCM is simply to get your data and models aligned and the rest will go easy to find the optimal set of decisions for a firm to make. I have been in numerous situations where different groups have different interpretations of the same data. This can range from factory performance information to projected demand. Projecting demand is particularly volatile when substantial asset decisions (production starts, purchases, capacity, etc.) need to be made.  Having owned models, one learns when the model results projecting the future state does not match the actual events in the future, you are the first to face the inquisition. The projected results are never published with the assumptions made or few people read them.

Interplay. There is a component of technology investment to improve organization decisions.  Making that is a leap of faith. One can identify the possible opportunity or similar success elsewhere, but the rest is conjecture (well-reasoned in some cases). If one is attempting to push the boundary of best in class, many landmines exist. In the early 1980s, I was a member of the LMS team to create an application with real-time data, rules, and models to handle the complexity of dispatch scheduling for factories that produce wafers. We were promoting the “fancy stuff”. A competitive approach at that time was a much simpler manual Kanban approach – “simpler, but more robust”.  Each side was driven by beliefs and there is always a component of survival.  Over time LMS proved to be a great success and this approach dominated dispatch scheduling in Wafer production. Early in the evolution of LMS “electronic Kanbans” were put into its logic. I witnessed many similar situations, for example, the use of optimization for central planning. In each case technology upset the social order.

Intermittent or sporadic events. The bane of all demand-supply networks from demand to supply are interruptions to manufacturing excursions. Success in managing these events requires a shift in orientation to risk management from optimal efficiency. For any large organization, this is impossible to do without quality models (these days called the digital twin).

Conclusion

In previous blogs, we contended COVID-19 was an operations management challenge.  Over the past six months, significant progress has been made in understanding COVID-19, reducing its health impact, and attempting to do this with a limited economic impact – using core principles from operations management. However, substantially different views exist on what should have been done and the best course forward. These differences relate to the following issues:

  1. Gaps in the data
  2. Same data and models, but different interpretations
  3. The interplay between policies and beliefs – upsetting the social order
  4. Intermittent or sporadic events

We demonstrate these issues cannot be solved simply with the application of analytical methods given the current data gaps and the inherent nature of the policy. These same challenges, albeit in a different environment, can impede the success of any investment in new technology to improve supply chain management decision making. A successful implementation requires strong leadership that can navigate the political terrain to build consensus supported by agents of change who harness community intelligence. If not, warring nations will undermine the effort. As Machiavelli observed: “There is no more delicate matter to take in hand, nor more dangerous to conduct, nor more doubtful of success than to step up as a leader in the introduction of changes.  For he who innovates will have for his enemies all those who are well off under the existing order of things and only lukewarm support in those who might be better off under the new.”