When we first saw the Snowballers’ dramatic results, we thought they were either lucky or genius at picking a lucrative use-case at the outset of their advanced analytics journey. We thought they were perhaps better than the rest in answering “What proof-of-concept AI use case should we pick.” It turns out their strength was rooted not so much in being adept at picking the right use-case, but enabling their enterprise to pursue any use-case.
Like all others, the Snowballers did not start off as AI luminaries. But they became adept at seamlessly turning BI-oriented hindsight into AI-driven foresight. By putting ML Ops at the top of the enterprise’s analytics value-chain, Snowballer managers removed the BI-AI boundary. Powerful AI-driven insights flowed to anywhere within the enterprise to where BI was already flowing — which essentially was everywhere.
Snowballer managers pursued AI differently than their contemporaries. Rather than picking the “right” AI use case, they enabled AI use-cases. Such action resembled what previous generations of managers did when confronted with game-changing General Purpose Technologies (GPT) like AI. Economists define a GPT as any innovation that becomes a springboard or platform for creating new innovations. Electricity is an excellent example of a GPT. Harvard University economist Robert Gordon sheds light on how this earlier GPT reconfigured competitive fitness:
“Introduction of electricity into American factories did not immediately raise productivity, simply because it merely replaced steam engines. It was only when bosses realized that electric motors allowed factories to be reorganized — dispensing with the need for machines to be close to a central power source — that productivity soared, as workflow improved, and new cheaper buildings could be used.”
The history of GPTs gave us an essential clue as to why our Snowballer clients were dramatically pulling ahead of the pack despite making seemingly simple changes. Snowballer managers implicitly saw AI as being most effective when it is operationally stitched into core business processes and those core processes are tweaked. They reorganized existing enterprise tasks and decision-making in simple ways to harness the power of AI.
For example, “Nike pushed the boundaries of real-time, analytically-driven advertising. Within seconds of a memorable shot, Nike delivered ads across sites and apps in 15 countries. Fans interacted with these moments, making them their own, and share them on social networks” (see here). From 2010 to 2020, Nike outperformed the S&P 500 by 230%.
In another example, Disney reports that “all our data and [predictive insights] are designed to help Disney cast members anticipate all your desires so they can give you an incredible experience. The goal of our tech team was to root out all the friction within the Disney World experience… data fuels a better experience” (see here). From 2010 to 2020, Disney outperformed the S&P 500 by 157%.
In contrast to the above B2C marketing darlings, B2B companies also comprise the roster of Snowballers. One such example is the large chemicals company Ecolab: “We collect operational data and leverage enterprise technology platforms to give us a view of our customers’ operations, which we then leverage towards more innovation, more solutions, better answers, and enduring relationships. Insights gleaned from this data have transformed Ecolab’s field service operations” (see here). As a result of this analytical transformation in a 97-year-old company, Ecolab outperformed the S&P 500 by 160% between 2010 and 2020.
The Snowballers we served are especially impressive because the majority were not digitally disruptive startups. Neither size nor age of the enterprise governed whether they emerged as Snowballers. Nowadays, every enterprise likes to say they are data-driven. The Snowballers, in contrast, have been model-driven.
This is true for small organizations as well as behemoths, for startups and incumbents. One of the CEOs we worked with, who leads the Biotech startup Trugenomix, put it eloquently: “Our knowledge of patients and our power to change patient outcomes must be glued together; if we don’t know the cause of why something happened, we can’t produce the effect we want.” This CEO leveraged the DeepCortex ML Ops platform to integrate hindsight analytics regarding patients’ health data with predictive models to produce unprecedented insights on each patient’s precise diagnosis, prognosis, and prescription.
Snowballer managers across the board were no different than this Biotech CEO (but different from the other managers, including the Elite-10) — they moved ML Ops up the analytical chain to ensure hindsight and foresight were seamlessly integrated to yield the strongest impact possible.
Placing ML Ops far up in the analytical chain makes a huge impact due to a hidden killer of business value: Lag. Lag is the time it takes between when a business action is performed and when the effect from that action is measurable by the business. In 1950, MIT professor Jay Forrester demonstrated how insidious lag effects could be on an organization by introducing a beer game to help managers simulate real-world supply-chain dynamics.
The supply chain for beer consists of a consumer, retailer, distributor, wholesaler, and manufacturer. Demand originates with the consumer. Because of the lag in terms of how frequently the consumer can restock their fridge, they purchase slightly more than they need when they see their fridge getting low and nice weather for a barbeque approaching. The retailer suddenly sees its stock levels decrease and places an order for more beer.
Since there is a lag between the distributor receiving the retailer’s order and getting additional cases from the wholesaler, the retailer does not see their replenishment for another two weeks. In the meantime, more consumers purchase beer, and the retailer’s stock continues to deplete. Failing to see any replenishment from the distributor for the previous order, the retailer doubles down and places an even larger order. These orders pile up at the distributor, who in turn places even larger orders with the wholesaler. Due to the lag time in increasing the manufacturer’s supply, the wholesaler sees a growing pile of unfilled orders. To restore equilibrium, the wholesaler also doubles down and places more orders on top of the previous orders that have yet to be fulfilled.
Finally, six weeks later, when the nice weather for a barbeque has passed, all the accrued orders begin getting passed down the chain too late. This results in a glut of beer at the retailer. To achieve equilibrium here, the retailer ceases to place any orders for the next several weeks, although normal consumer demand is still present. The manufacturer sees zero orders coming in and assumes consumer demand has dried up. The manufacturer halts the manufacturing lines, and the whole cycle of whiplash begins anew.
Every actor in the chain is whiplashed due to lag effects across the chain. Each actor fails to make effective management decisions to optimize profits:
Just as there are lag effects between companies, there are lag effects within companies. Lag exists between employees as information works its way from customer-touchpoints to the back-of-the-house. Or as information flows from the front-lines to the executive managers. Internal supply chains translate every activity every employee does to eventual profit. Wherever there are chains of information flow, there are lag effects. Wherever there are lag effects, there is an erosion of profitability and customer satisfaction.
Lag makes adjusting any lever of the business that drives profitability nearly impossible — it makes adjusting prices, or adjusting purchase volume, or marketing spend, or preventative maintenance schedules factory machines, or department budget allocations, or whatever — a futile exercise.
Think of it like this: Imagine there is a 5 minute lag time between when you adjusted the hot water knob on the faucet and when you felt the temperature change. Imagine how many times you would need to turn the knob to get the temperature just right to wash your hands. Lag effects make getting the plethora of knobs-and-dials that comprise a modern business just right tough.
Now imagine you are in a race against a competitor to see who can finish washing their hands first. There are two ways to wage this competition. First, you can reduce the lag between when a knob is adjusted and when its effect is measured. This enables a more agile market-driven business.
Second, you can build a predictive model from your historical data on previous knob-turns and water temperature outcomes. That model gives you the ability to predict the magnitude of knob-turn you need to get the water temperature you want. This enables a predictive market-driving business.
Becoming competent at being either market-driven or market-driving provides a leg up on the competition. Becoming competent at both disrupts the competition.
The Snowballers showed us over the years that becoming effective at both creates a deep moat around the business and confers an enduring competitive advantage. Through ML Ops, the Snowballers instituted the ability to not only rapidly build and operationalize predictive models but derive predictive AI models as a natural extension of their BI pipelines. BI enabled them to be market-driven, while AI enabled them to be market-driving. Pervasive ML Ops enabled them to be both.
Business Intelligence shows the as-is state of the business. Advanced analytics shows the will-be state of the business. The “will-be” state cannot be predicted without knowing the “as-is” state. By integrating BI and AI with a common ML Ops platform, the Snowballers achieved a continuous flow of high-speed optimal decision making that circumvented lag. Snowballers imbued human actors and robotic IT systems with the hindsight-and-foresight required to take the right actions at the right intensity at the right time — all the time.
A small initial advantage snowballs into a significant (and sometimes, unassailable) long-term advantage due to self-reinforcing loops. In a self-reinforcing loop, chains of cause-and-effect double-back on themselves to create a flywheel of growth. Advanced analytics and BI seamlessly integrated into operational business processes give rise to natural self-reinforcing loops: