A couple weeks ago, Parker Avery’s Chief Analytics Officer Sam Iosevich and Partner Amanda Astrologo participated in a CGT webinar where they discussed the benefits of a centralized demand signal.  Over the last few months, Parker Avery has helped several clients realize tremendous improvements from implementing and adopting this new way of approaching and using demand forecasting.  However, many of the questions received during the webinar made us realize that the overall concept is not clear to those who haven’t been “in the weeds” with this new way of operating.

Demystifying the Centralized Demand Signal

This week, we demystify the concept of a centralized demand signal, explaining it in plain terms and answering the question:

“What exactly is a centralized demand signal?”

A centralized demand signal is a single, holistic view of demand across the enterprise. Sounds great, but what does that really mean? Let’s take a quick step back.

A dysfunctional history.

Traditionally, retailers and consumer brand companies (CPGs) operated in functional silos. What this entailed is that product development, supply chain, merchandising/category management, store operations, marketing, finance, and other departments each came up with their own version of what they expected consumers to purchase from the company. They then made and executed plans according to each department’s expectations. These plans affected planning, pricing, fulfillment and operations.  Today, most companies continue in this manner: buying product to one signal, planning pricing to another, fulfilling to a third, and planning labor and transportation to yet another.

These individual forecasts may include not just how many items, but in what geographies, and in the last couple of decades: in which channels. For added fun, when online shopping started becoming mainstream, most retailers added another (yes, separate) functional area: e-commerce. And then of course, mobile crashed the party, in all its glorified ubiquity.

Only in the last few years, since the quick evolution into omnichannel (or unified commerce if you prefer) have the store and digital teams figured out that consumers don’t shop in different channels. Consumers shop with the brand.  This epiphany led to the notion of a “seamless customer experience,” despite multiple ways of presenting products and conducting transactions. Some could say that the move to unified commerce was the spark that led to the notion of functional collaboration, and—here it comes: a centralized demand signal.

Let’s get into the details.

A centralized demand signal is a combination of a stable, unbiased demand forecast across the geography, product, and time hierarchies. Admittedly a mouthful. Let’s unravel it.

The stability of the forecast means that there are no extreme projections across the geography, product, and time hierarchies. This allows different functions who operate at different levels of the geography and product hierarchy at different time horizons and granularity to utilize the same consolidated forecast.

An unbiased forecast means the forecast is not consistently wrong in one direction. Meaning, consistent over or under estimation of demand often leads to out-of-stocks and shrink through excess inventory. While accuracy is important, an unbiased forecast is a critical starting point.

Further, a centralized demand signal starts with a true unconstrained view of demand. We use “unconstrained” to describe what we think the true demand is, assuming unbounded inventory. Then adjustments are made based on constraints related to inventory, supply chain capabilities, resource availability, or allocation decisions (channel or location). For example, a retailer selling iPhone 12s, will only get so many of each model and popular colors. Similarly, a department store receives a limited assortment and supply of certain brands (e.g., Tory Burch, Ralph Lauren). On the other hand, a grocery store can find multiple suppliers of whole chickens and produce, but they may be constrained on labor to process/prep them for the stores. The initially unconstrained demand is adjusted based on those types of constraints.

Where does the signal come from?

Here’s where the deep mathematics and science comes into play. The signal is created using a company’s historical data and combining it with demand driver information such as digital spend, macroeconomic variables, and pricing, as well as downstream sales data (sometimes purchased) and even data from highly disruptive events—for example, a global pandemic.

Now let’s add artificial intelligence that continues to monitor the right level to model and focused machine learning that improves the forecast as it bounces the science-and-lots-of-good-data-based estimate against actual outcomes.

Good signal. One lots of people across the company can trust. Especially when it’s far closer to actual numbers than most companies have ever imagined. Sit. Stay. Good signal.

How does it help with collaboration?

No machine is perfect—although Parker Avery has helped some clients realize actual results that come pretty close to the forecast. Now that there’s a demand signal that’s universally trusted across the organization, the magic of collaboration comes in to play when that signal is informed by functional insight from across the organization. Put simply, it contains inputs from all functional areas and aligns the organization to a common view of demand.

Each functional area knows the nuances that need to be incorporated into the forecast, and adjustments should be made to accommodate those idiosyncrasies. However, this should not be done in isolation. Working transparently across functional departments ensures all stakeholders understand where changes must be made and more importantly—why those changes may be required.

An example of this could be supply disruptions due to container shipping issues (remember the Suez Canal incident?) or one-off promotional events.

Once changes are understood and incorporated into the forecast, the company can confidently operate under a unified, centralized demand signal. Each functional area can perform their activities marching to the same proverbial drumbeat. This forecast ties to category and item promotional and pricing decisions that drive replenishment and store operation decisions. Each function consumes the elements and level of detail required for their actions and decisions.

Let’s talk benefits.

The most important benefit is agility. In a market facing global disruptions combined with the intense acceleration of omnichannel, agility becomes increasingly important. It is extremely difficult to adjust to disruption when operating on a dozen or more demand signals across the enterprise. A centralized demand signal allows you react quickly to an ever-changing landscape.

Other benefits associated with use of a centralized demand signal include reporting simplification, improved communication throughout the company, better understanding of the needs of other functional areas through more standardized collaboration, as well as better data management and data governance.

With a centralized demand signal, there is no questioning what one system uses for data vs. another or if the assumptions driving the forecasts are different. Those conversations are mitigated or more likely eliminated, and the organization can focus on more strategic activities.

Further, the combination of AI forecasting and human insight is consistently evolving. As a result, more mundane and predictable planning tasks can be automated, allowing employees to concentrate on higher value-added tasks. The transparency of the inputs to the AI forecasting process is key for employees to be able to concentrate on the attributes not covered by the forecast.

How do we begin?

The first step in any analytics journey is to understand the state of your data. Having good, reliable historical data is absolutely critical to a good forecast. Unfortunately, this is not the sexy, magical AI/machine-learning stuff of War Games. But without a solid starting point of clean, historical data, even the world’s most advanced models will not produce a reliable forecast.

The second step is to demystify AI for the organization. It is not the human against the machine; rather, it is the human working with the machine to provide a superior result. AI should focus the human on more value-added tasks; this is a benefit to the employee and the organization.

If you have any questions or would like to learn more about Parker Avery’s approach to centralized demand forecasting, our analytics services, or any of our other retail and consumer goods consulting services, please don’t hesitate to contact us.

Watch the Webinar
Contact Parker Avery

For additional Parker Avery perspectives, please visit our insights page.