Summary

Over the past two months I have been making the case that just “data driven” is a road to disaster. Data is a component of the effective decision-making processes in operations management (OM), but far from the only component. If you are thinking “machine learning and AI” will save you from data disasters – think again as the pandemic behavior is playing havoc with machine learning models (a lesson those using Box Jenkins ARIMA models learned in the 1970s).

Our weird behavior during the pandemic is messing with AI models

Without other components disaster is inevitable – of particular import is handling various objectives simultaneously – a skill set any effective factory or supply chain direction has in abundance, but so far in short supply in responding to COVID-19.  The importance of OM in effective response to COVID-19 is gaining traction as we see in the webinar sponsored by the University of Cincinnati titled “Decision Making Under COVID-19: Planning and Recovery.” Last, there is a growing thought process that there is not that many COVID-19 cases and for the most part it is not that serious.  In many ways this is a direct result of OM failures by leadership – especially the inability to focus on multiple objectives at once.  A close friend of mine is a best in class pulmonologist on the front lines since the initial outbreak in New Rochelle, NY.  He has patients of all ages and if you get sick enough to spend quality time with him – it will not be a fun three weeks.


Introduction

Over the past two months I have been making the case that just “data driven” is a road to disaster.

Lessons for COVID-19 and Supply Chain Management Models

Statistical Forensics – The Danger of Being Data and Not Operations Management Driven

Data is a component of the effective decision-making processes well established in operations management (OM), but far from the only component.  Other components include detail working knowledge of the fundamental structures of the network, understanding it is a connected network, models to support various decision levels (which includes bottoms up modeling), predictive methods, community intelligence, and the ability to attend to simultaneous objectives at the same time.  The last is a critical skill set effective factory and supply chain directors have in abundance.  This blog provides some examples of limitations and the emergence of OM as the critical skill in the responding to COVID-19 by pointing the reader to an upcoming Webinar sponsored by the University of Cincinnati titled  “Decision Making Under COVID-19: Planning and Recovery.

Examples of Limitations

We have seen the limitation of data drive in the COVID-19 response over March, April, and May.  Those responsible for responding (government leaders, epidemiologists, and public health) have succumb to some standard traps, for example:

  1. We need data, any data, to post and watch- without a full understanding of its limitations, intricacies, and ability to capture and have a common accepted definition.
    1. Trump Suggests Virus Death Count Is Inflated. Most Experts Doubt It
  2. Following a misguided KISS (keep it simple stupid) principal by aggregating the data or failing to adjust it.
    1. D.C. Test Counting Error Leaves Epidemiologists ‘Really Baffled’
    2. Why Do Countries’ COVID-19 Death Rates Vary So Much?
  3. In Delaware only in the last month has it posted COVID-19 data (cases, hospitalizations, and deaths) per capita when comparing counties (before it was totals) – frankly, a rookie mistake.
  4. Assuming all of the answers could be found in the data as opposed to reasoning from first principals.
    1. This lead Delaware to ban drivers from PA (which it shares a large border and is economically integrated with) except to work in delivering essential services while a COVID-19 fire storm in its southern most county (Sussex) exploded (150 miles away from Pennsylvania).
    2. It took CDC at least two critical months to determine that masks are critical to control the spread. Most of the spread is from respiration (seeing your breath on a cold day).  While folks were wiping down grocery store carts and wearing gloves, they are traveling around the stores without masks.
  5. Inability to focus on multiple objectives simultaneously, for example failure to attend to hunger for our youth. The Delaware COVID-19 data site has ZERO information tracking this.  There is no logic in keeping a child COVID-19 free why they remain malnourished or missing critical education.
    1. Emergency Child Hunger Program Is Far Behind on Rollout

The CDC failure on masks, not only permitted preventable spread, but has undermined confidence in government which has generated views that COVID-19 isn’t that bad, only a few people pass away, everyone else gets better in a few days, we need everyone to get it to create herd immunity, and government/epidemiologist/public health officials exaggerated how dangerous COVID-19 is (an example of trap # 2-failure to use models appropriately).

The failure on hunger and education sends the message leadership is overwhelmed – again undermining confidence.  The list can go on and on.

Webinar: Decision Making Under COVID-19: Planning & Recovery”

The application of operations management methods to help guide COVID-19 response is gaining momentum. I want to point readers to the upcoming Webinar series sponsored by the University of Cincinnati under the direction of Prof. Michael Fry. The first in the summer series is on June 1 “Analytics Leadership in Uncertain Times.

One session of particular interest is “Analytics During a Crisis” – Four seasoned leaders of analytics in the CPG, Retail, and Food industries discuss how they empowered their organizations to solve real-time problems related to the COVID-19 pandemic. Learn how agility counts when facing significant disruption and how inspirational leadership matters.

The second in the series on June 15th is specific to OM and COVID-19 titled “Decision Making Under COVID-19: Planning & Recovery.

There are three sessions:

Session 1 – “COVID-19 Scratch Models to Support Local Decisions” Edward H. Kaplan, William N. and Marie A. Beach Professor of Operations Research, Professor of Public Health, Professor of Engineering, Yale School of Management, Yale University

Abstract: I was appointed to Yale University’s COVID-19 advisory committee to provide analyses supporting university decisions during the early weeks of the SARS-CoV-2 outbreak. This work expanded in response to requests from the Yale New Haven Hospital and the State of Connecticut for help. Much of this work relied on scratch modeling, that is, models created from scratch in real time. Applications to date include determining crowd-size restrictions on events, hospital surge planning, university shutdown and restart timing decisions, designing viral testing programs, and environmental monitoring by testing sludge from the local wastewater treatment plant. I will describe the problems faced, types of models developed, and advice offered during real-time response to the COVID-19 crisis at the local level.

Session 2 – “How to Prepare for Ramping Up? Riding the COVID-19 Wave in the Next Six Months” Jan C. Fransoo, Professor of Operations Management and Logistics, Kuehne Logistics University

Abstract: Extending our successful dynamic models that we deployed 11 years ago during the credit crisis, we model the production and market lockdowns caused by the COVID-19 crisis. We estimate the supply chain dynamics that we may see unfold over the next few months. Our results show that inventory dynamics may be very large, caused by dramatic drops in demand. Regardless of how the market recovery will evolve, we demonstrate the criticality of monitoring cumulative supply chain inventory and market demand. For companies upstream in the supply chain, the impact of the inventory evolution is much stronger than the exact details of the market recovery.

Session 3 “Interpreting Predictive Models Related To COVID-19”

Three experts discuss the challenges related to interpreting predictive models subject to extreme uncertainty. All analytical models are subject to the limitations of existing data and underlying assumptions. Models related to COVID-19 are especially challenging to interpret due to the lack of historical data and the inherent uncertainty of human behavior during these unprecedented times. Learn from experts on how best to communicate model results and insights to decision makers when models are subject to high degrees of uncertainty.

All three sessions will promote and shed considerable light on the value of Operations management with regards to COVID-19 and managing demand supply networks overall.

Closing Note

A close friend of mine for over 50 years is a best in class practicing pulmonologist with a strong background in critical care and infectious disease who came out of semi-retirement to treat COVID-19 patients. He noted COVID-19 is not just a disease of the old.

These Athletes Had the Coronavirus. Will They Ever Be the Same?

If you get a case of COVID-19 where you need a hospital and he need to treat you – you are in for an unpleasant experience for 3 weeks plus a long recovery.   The skill set to navigate serious COVID-19 patients back to health is in limited supply – if too many show up at the same time, the Inn is full.   He was on the front lines in NYC. The “front” can show up any place where there is a vector in and failure to be observant. 

Coronavirus Ravaged a Choir. But Isolation Helped Contain It.

Enjoyed this post? Subscribe or follow Arkieva on LinkedinTwitter, and Facebook for blog updates.