Mitigating unknowns in an age of business uncertainty

Kurt Marko Profile picture for user kmarko July 6, 2021
A recent column from Salesforce's Peter Coffee and the death of Donald Rumsfeld leads to a rumination on the epistemology of business information and decision inputs and the development of a theory on how to thrive in conditions of extreme uncertainty.


The last 18 months have been a time of unparalleled (at least in the post-WWII era) uncertainty for businesses and society writ large, a situation that has created great peril and tremendous opportunity. Although it seems that the worst of the chaos is over, the pandemic bounceback might prove as disruptive as the initial shock in the same way that the damage from a tsunami doesn't all come from the initial wave, but also from the riptide of water rushing back to sea.

Having navigated the initial disruption and prolonged uncertainty, businesses face a new set of post-pandemic "unknowables," as outlined by Peter Coffee from Salesforce. These are:

  1. How rapidly and to what extent "will people return to pre-pandemic behaviors?"
  2. Will changes, for example, increased use of online ordering and delivery or curbside pickup, catalyzed by the pandemic, be lasting, temporary, or recalibrate to a new behavioral mix?
  3. "What new behaviors will emerge" from pandemic-caused adaptations and what types of business models, or conversely, "creative destruction of old business models," will result?

Peter's column on uncertainty reminded me of the famous quote from the recently deceased former U.S. Defense Secretary Donald Rumsfeld, who famously responded to a question about Iraq War tactics with the following:

There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we know we don't know. But there are also unknown unknowns. There are things we don't know we don't know.

While the BBC obituary claims that Rumsfeld was widely mocked for his answer, the ridicule was a combination of reporters feeling slighted and their lack of appreciation for the epistemological truths embedded in Rumsfeld's statement. Indeed, this framing of uncertainty is useful for understanding how best to address Peter's questions and cope with the pandemic-fueled migration of business processes to online, automated platforms.

Prioritizing unknowns

As I indicated in a Twitter response to Coffee's column, using 'Rumsfeldian' framing, there are two types of unknowns:

  1. Factors we can identify, but not predict (at least well) or accurately quantify. These are the "known unknowns."
  2. Events that we can reasonably expect to happen, but can't identify, categorize, quantify or temporally predict. These are the "unknown unknowns."

Elaborating on my Tweet, business leaders can address and mitigate each of these by planning and design, but in very different ways:

  1. Narrow the range of uncertainty of type-1 unknowns via projects to study factors known to pose risks should the organization maintain the status quo, but can be better understood through focused technical, market and economic research. These might include an upstart competitor with a different business model that customers find more cost-effective, convenient, or of greater utility. These often rely on disruptive innovations that exploit new technologies such as cloud computing, machine learning or mobile devices to provide a product or service in ways and at costs incumbent providers can't match.
  2. Improve the ability to react to and capitalize on unpredictable events, i.e. the type-2 unknowns, by creating an adaptable organization by exploiting technology and software-defined and managed business processes. Such systems that can withstand surprise disruptions, disasters, attacks or cascading mistakes are called antifragile. In contrast to robust systems that resist failures (a nuclear submarine or the Golden Gate Bridge) and resilient systems that quickly recover from failures or disruptions (the Internet, the human immune system), anti-fragile systems can accommodate or adjust to changes for which they weren't designed or even knowable at during their conception. Examples of unknowable adaptability we saw during the pandemic included:
    1. The rapid adoption of existing but not universally deployed work-from-home (WFH) technologies including video conferencing, SD-WAN and online collaboration discussion and project management tools (Slack, Asana, Basecamp, etc.).
    2. Aggressive use of cloud infrastructure (IaaS) and applications (SaaS) that can be quickly deployed, scaled and reconfigured to meet changing demand or employee restrictions.


Examples of preparing for known unknowns

As I detailed last summer, an overarching consequence of the pandemic response was the accelerated enterprise adoption of cloud services. Although the initial increases were reactionary, to accommodate unpredicted changes in resource usage and office closures, over time business leaders shifted to strategic considerations about the future role of private versus public cloud infrastructure and the risks of concentrating critical applications and data at a single provider.

Cloud architectural planning usually raises several type-1 (known) unknowns that must be analyzed and addressed to avoid deleterious surprises. Key uncertainties concern the costs and benefits of a simple hybrid (private plus one public) versus multi-cloud (private plus multiple public) cloud designs. Surveys show that most organizations adopt a multi-cloud strategy for one of several reasons:

  • The option to choose among the best available services from several vendors across application and infrastructure categories.
  • To hedge costs for undifferentiated services by moving workloads to the cheapest provider.
  • To improve reliability by distributing applications and data across multiple providers in different regions.

Unfortunately, a multi-cloud strategy creates known unknowns, such as:

  • Unexpectedly large bills from unconstrained usage exacerbated by the low friction of instantiating and scaling cloud resources and inadequate oversight of cloud resource usage and costs.
  • Rapidly increasing costs to move data between cloud providers which surprises organizations that don't understand how cloud egress costs create friction and data gravity. The ideal vision of a seamless multi-cloud architecture becomes much less attractive when factoring in all the resistive forces impeding data and workload movement between environments. Indeed, such data gravity is a reason many organizations adopt a cloud-agnostic data fabric such as NetApp or HPE Cloud Volumes (separate products), VMware vSAN or similar products. Similarly, multi-cloud storage friction presents an opportunity for third parties like Box, Dropbox, Flexify and others to build intermediary services that can provide data to users, applications and cloud infrastructure when needed, and reduce storage and egress costs.

Wide swings in customer demand for products and services were another consequence of pandemic uncertainty that, one year on, are confounding traditional means of adjusting to seasonal and systemic variations. As business returns to some semblance of normal, companies in retail, logistics, travel and hospitality face unprecedented surges and chaotic swings in demand that they struggle to anticipate. Consequently, more organizations seek to mitigate these known unknowns through the aggressive collection and predictive analysis of relevant data using sophisticated machine learning models.

A recent Wall Street Journal article detailed several examples of companies using "real-time data to feed predictive software models, including live website activity or online and mobile searches." For example, travel booking agency Hopper, but the company's head of price intelligence recognizes that 2020 isn't representative of current demand curves. Although Hopper has compiled an archive of "several trillion seat prices" collected over the past few years, it is shifting from historical and seasonal trend lines to more real-time predictive analytics.

Walmart has also eschewed annual and seasonal trending in favor of ML models that "crunch years of internal sales, pricing and market share data, as well as industry reports, economic forecasts and broader retail trends." According to a Walmart spokesman, the analytics effort has “helped us normalize the data and more accurately perform demand forecasting.” Indeed, a 2020 Forrester survey "of 300 enterprise decision-makers" around the world found that 64% of respondents are "making operational data readily accessible" and 43% are "undertaking a strategic initiative to identify use cases for AI/ML" and developing internal AI/ML applications.


My take

The past 18 months have provided a stark lesson in the importance of fortifying an organization to tackle the unknown. Much like cities in California prepare for the eventual Big One and those in coastal Florida are ready for the eye of a Cat-3, business leaders must have a two-pronged strategy to survive and thrive by being able to

  1. Respond to known unknowns such as disruptive innovators and rapid shifts in consumer behavior via rigorous planning and adaptable and automatable processes.
  2. Confront unknown unknowns by building an anti-fragile organization that can bend, but not break in response to disasters and other unexpected business calamities.
A grey colored placeholder image