At this point in time, the words “we are at war” have been spoken by several national leaders. But what does this actually mean? In any war, we have the front lines of battle. Typically, that involves defense personnel in a battlefield located on a country’s border or abroad. In a war against a highly contagious virus, battles are fought in our hospitals and health care personnel are our warriors. But wars are also about logistics. It is imperative that front line workers get the right supplies that preserve their health and provide them with the right weapons to fight. At the same time, as war affected populations go underground (or quarantine in this case), we need to ensure that the most vulnerable among us are taken care of and provided with essentials. Managing the logistics is easier said than done.
The Blitzscaling Challenge for Logistics
As Jose Andres writes in his inspirational NYTimes article, https://www.nytimes.com/2020/03/22/opinion/restaurants-coronavirus-food-aid.html
Solve the informational and logistical challenge: Matching demand and supply — by getting food to the people who need it most — is even more challenging than cooking in a crisis. Distribution is the Achilles’ heel of any disaster response.
These issues are severe when supply is limited and demand is rapidly evolving. Doing these effectively (covering as many people as possible) while being efficient (cost/time) is critical. The exact same issue applies to providing Personal Protective Equipment (PPE) to health care workers. Here is a statement from a volunteer in PPE procurement and delivery:
What I saw was confusion about how to gather and deliver supplies. It’s a local problem. The hospitals don’t want 1000 volunteers or donations walking up to their door. They are swamped, and it’s an infectious area. There needs to be vetting and batching, and single point of contact with the facility that is low friction for them. Organizing this in every metro area a huge challenge. Too many people want to help, but don’t know how, etc. Also the facilities have varying standards for what they will accept. If you give them material they can’t use, they will waste it, when someone else might have wanted it. We need a Process to filter the confusion, and maximize what they can actually deliver in a short time.
Thus, the need of the hour is to be able to collect data, sift through it, analyze it and execute in a way that maximizes our objectives but do it all rapidly, essentially the notion of blitzscaling : https://hbr.org/2016/04/blitzscaling . In fact, Reid Hoffman, the originator of this term makes a close association between this term and the use of blitzkrieg tactics in World War II, which is where we will head next.
Some Comparisons with History
The modern field of Operations Research arose during World War II. In that era, Operations Research was defined as “a scientific method of providing executive departments with a quantitative basis for decisions regarding the operations under their control”. Other names for it included operational analysis and quantitative management. Sound familiar? Although the term Operations Research continues to be used in optimization and supply chain circles, the term data science which includes statistics and machine learning has captured popular imagination. The goals of data science are exactly the same as that of operations research: to use the data and technology available in a scientific way to solve challenging problems. And Covid19 presents us with a uniquely challenging problem, particularly for logistics.
Conventional wars involve humans on both sides. Human beings care about casualties and the battlefield is near equal in terms of how fast the adversaries can move. Even in that context, minor differences in capabilities can have huge impact on the outcome. The war on Covid19 is staggeringly imbalanced. The virus moves fast (contagion), in stealth (asymptomatic spread), and scrambles our ability to track (long incubation period). This is a war unlike one we have ever seen and a logistics response involving blitzscaling is pretty much essential.
Covid19 meets the Gig Marketplace
For a variety of non-controversial reasons (and I will only stick to them), dependence on governments alone is inadequate in a crisis like this. A disproportionate amount of supplies to be transported in this war are medical supplies such as PPE, not necessarily the government’s forte. Further, strategic government stockpiles of such supplies do not seem to have been calibrated for a pandemic of this scale. However, it turns out that for products such as N95 masks, private individuals and businesses have inventory available within their home or facilities. In times like these, they are more than willing to donate their supplies to hospitals and clinics. The need of the hour then, is a marketplace that can facilitate the transfer of supplies from the donors to recipients. This is not an ordinary challenge. There are several obstacles:
- Creating an online portal than can accept information from donors and recipients
- Matching recipients to the right donors given geographic distances (which affect speed and transportation cost) and giving donors the right information about their matches
- Handling large donors through a multi-step supply chain that may involve consolidation and cross-docking locations
- Ensuring that small donors do not have multiple touch points or channels to deal with when making these donations
- Partnering with the right transportation providers, whether large carriers or crowdsourced individuals in cars, to ensure speedy and reliable delivery
- Trying to do all of the above for a nationwide network in days (not weeks or months). This is Blitzscaling 2.0!
Communication and cloud technology has had a big impact in putting together the backbone of such an operation. However, the creation of a “process” to make this happen efficiently and at scale requires more than just technology. For example, matching of supply and demand is often done manually and as the number of donors and / or recipients increase, more volunteers are needed, increasing the level of complexity. Matching errors may ensue, leading to the kind of chaos referred to in the quotes above. Data Science in a cloud computing environment can be leveraged effectively to address these issues.
A catalog of my experiences
Over the past 16 days (or 16 Corona years if you prefer!). I have had the privilege of interacting with two volunteer efforts for PPE mobilization:
#GetUsPPE - Getting Protective Equipment to our Healthcare Heroes
Please join us today in helping to find, make, and bring Personal Protective Equipemnt (PPE) to our healthcare heroes…
Mask-Match is helping match front line healthcare workers with free N95 masks that people have in their homes.
Both these organizations are doing yeoman service to our society and composed entirely of volunteers. However, they face all the challenges I described earlier. My enthusiastic team of tech and data science volunteers has deployed the first version of a matching algorithm that minimizes transportation costs for donors while maximizing demand fulfillment for recipients such as hospitals and clinics. We are beginning trials at https://getusppe.org/ and hoping to be an integral part of their workflow. Our job is not done as the network is continually expanding and growing more complex with multidimensional objectives. Our larger goal with this initiative is to democratize these models and algorithms for any organization involved in non-profit and humanitarian efforts.
The adage that “crisis leads to innovation” is true for this pandemic as well. The blitzscaling of volunteer organizations using technology and data science to address the challenges of this pandemic is a template for the world to follow. But the real discovery here is the incredible capacity of human beings, from different walks of life, to come together and cooperatively tackle huge problems. That fact alone fills me with tremendous optimism for the future of human civilization.
Acknowledgements: I would like to thank Rohit Jacob, Neeraj Bhagchandani and Arun Rao for putting together the infrastructure and code for automated matching. Thanks also to Shuhan He, Charlotte Lee and Dan Lurie of getusppe for giving us this opportunity to contribute.