The question of accuracy vs. agility has been a long debated one in the CPG demand planning world. Usually accuracy and agility are seen at loggerheads with each other, especially during turbulent times marked by heavy promotional and competitor activity, recessions, weather calamities and pandemics such as COVID-19.
This is due in part, to the inability of current demand planning processes and systems to anticipate and proactively respond to volatility. Many systems and processes are not learning enough to adapt to increasing complexity and volatility. This leads to demand plans and forecasts being grossly off during critical times.
As a result, it is not uncommon for planning and execution teams (on both the demand and supply side) to anecdotally take a pause on the regular planning process and systems and switch to “agile” execution – expediting orders and production runs, rebalancing inventory, etc. During these times, more focus is given to cross-functional “agile” war rooms that do anything necessary to overcome the situation before moving back to normal cycles.
However, we at Samya feel that accuracy and agility don’t have to be at opposition. Let’s take a re-look at each aspect and how they can actually come together in a new way, enabled by the right mind-set, business processes and context-aware AI/ML technology.
When most people talk about agility, they talk about agility of execution – post the occurrence of an event/exception and how fast one can respond to that exception. This is what I like to call “reaction agility” of execution and that’s where many of the terms like “time to react” and “time to recover” have gained prominence.
Along with that is the realization that reaction agility comes with its own costs. To have reaction agility, not only does one need the right visibility, but also the ability to expedite orders and shipments, rebalance/transfer inventory, make exception deliveries and production runs, etc. All of this increases the cost of the supply chain. Although many practitioners and academics are studying the better design of supply chains and systems for reaction agility, it is still “reactive” in nature since it responds to a situation that has already occurred.
While reaction agility in execution is very important, anticipation agility in demand planning is also possible. What if, instead of reacting to an already occurred event (e.g. a demand upsurge down-surge, an out-of-stock event, etc.), one could anticipate the event well in advance (days, weeks, months) to either mitigate or leverage the impact of that event. That would be true agility borne out of anticipation.
This kind of agility embraces future change. While many S&OP processes are focused on “what has changed”, why not focus on “what could change in the future”?
If you take an example from the agile software development world, the agile manifesto was created to embrace change in requirements from customers and users giving rise to a whole movement with principles such as user interactions over documentation, shorter sprints of development, releasing working code as soon as possible, test driven development, CICD, etc.
In such a methodology the change happens but is easily accommodated as part of the process and not as a shock to the system. Bringing a similar line of thinking to the demand planning world requires planning and forecasting cycles to become more agile in anticipation.
Traditional wisdom in the demand planning world is that to be accurate you need more time for data points to develop, more time for communication and consensus building, and therefore more time to set an accurate forecast.
This mindset is strengthened by the practice of long monthly planning cycles, delayed system updates, long lead times of information communication from one function to another and has now become a self-reinforcing truth. In many cases, the existing forecasting engines and tools are not able to adapt since they were designed to do monthly forecasting at a certain level of granularity. And moreover, they are not able to adapt to go from monthly to weekly and shorter frequency cycles.
With this comes the conclusion that accuracy and agility cannot go hand in hand. Especially in times of increased volatility since in those times the forecasts are way-off and execution is what matters. This is true when the forecasting and planning cycles have not been designed for anticipation. However, if agility is seen in an anticipatory perspective as discussed above then a new picture can emerge.
For this let’s take a re-look at forecast accuracy. Accuracy has two aspects:
1) First by definition of the term, accuracy refers to how close the forecast is to the actuals, and
2) Second, how consistent is the forecast accuracy over a period of time.
While many organizations focus on accuracy, it is also important to focus on the consistency of forecast accuracy. For certain item-location combinations (at lower levels), even though the operational forecast accuracy might be high, the forecast accuracy consistency might be low. It is not uncommon to see operational forecast accuracies varying anywhere between two to six percentage points during the course of the year for many item-location combinations in the CPG industry.
However, if anticipation agility is infused into demand planning and forecasting cycles then it can help improve both forecast accuracy and forecast consistency.
With AI and Machine Learning, empowered by the right business context and human intelligence, demand planning and forecasting teams can be equipped with the ability to anticipate future changes or exceptions, empowering them with true agility that helps them be better prepared and improves both accuracy and the consistency of accuracy.
AI/ML can enable agile, interconnected and scalable forecasting with the following elements:
Agility and accuracy don’t necessarily need to be at logger heads. In-fact, anticipatory agility enabled by AI and Machine Learning can move the accuracy frontier forward in terms of validity and consistency driving significant business value.
Welcome to agile demand planning. Our engine blends intelligent cognitive automation and collaborative workflows to spot human bias, create transparency, and uncover previously hidden probabilities that will impact future inventory, promotions, and sales.