The landscape of forecasting is undergoing a significant transformation, driven by the advent of much smarter forecasting engines that utilize artificial intelligence (AI) and machine learning and the accessibility of structured “demand driver” data (for example, promotions, weather, or events). These innovations are dramatically improving the forecasting accuracy of planning systems, enabling them to achieve, and frequently surpass, the accuracy levels of human planners.
At the same time, the need for more granular forecasting has greatly expanded the volume of forecasts required. For example, companies are now looking to identify daily rather than weekly or monthly demand and to plan for individual stores instead of just distribution centers. Accordingly, the number of forecasts is exploding at a rate faster than can be handled manually.
Based on these trends, more and more companies aim to change their planning paradigm and strive for what we call “touchless planning.” Touchless planning is an innovative approach characterized by a high degree of automation in the planning process that reduces the need for human intervention. Instead of planners making every decision at the moment of truth—as in traditional human-led planning—planners now guide the system to make decisions on their behalf. As a consequence, the number of decisions planners are actively involved in is dropping massively, and their daily tasks and overall workflows are changing drastically (see Figure 1).
FIGURE 1: Traditional human-led planning vs. touchless planning
This shift not only streamlines the planning process but also allows planners to redirect their efforts toward more strategic activities and scenarios. By focusing on these areas, organizations can achieve significant cost reductions, boost their operational efficiency, and improve overall customer satisfaction.
However, touchless planning does require significant changes, and therefore should be implemented very carefully and thoughtfully. For touchless planning to be successful and efficient, planners need to trust the system and avoid the temptation to make unnecessary or wrong interventions that harm forecasting accuracy (and waste their time). Rather, planners should focus only on those areas where their input matters most, such as when planners have valuable information not available to the system (for example, information on local events, commercial insights from key customers, fashion trends, and competitor actions, to name just a few). In turn, where they are not adding value, they should let the machine take over.
Companies need to make sure they educate their planners about the new process and the key elements involved. In this article, we will outline how to gain planners’ trust in touchless planning and the most common challenges and key success factors.
Elements of touchless planning
Central to any touchless planning process are six elements: high-fidelity input data, right granularity, advanced AI engine, high-accuracy results, explainable forecast, and proactive feedback mechanisms (see Figure 2). We describe these elements in detail, as it is essential to understand how they interact and reinforce each other within the touchless planning ecosystem.
FIGURE 2: Periodic Table of elements to build trust in touchless planning
High-fidelity input data: Effective touchless planning requires complete, rich, and accurate data as an input; this includes both historical and future-oriented information. Historical data includes past sales, prior promotions, previous supply, and capacity constraints or stock outs that are essential to identify causal relationships and to quantify demand drivers. Often, cleaning and preparing this data involves a meticulous and labor-intensive process. For forecasting the future demand, forward-looking information must be available on drivers such as planned prices, upcoming retailer listings, and intended promotions.
Right granularity: The granularity of the input data is crucial, as it must provide precise information across various dimensions (for example, time, market, product, and product life cycle). The data input needs to be provided at the level at which it impacts demand. If the granularity of the input is not at the right detail, the AI engine is not able to learn from the data to forecast effectively. If you go too low level, the data is noisy and sparse, and if you go too high level, accuracy is lost in later disaggregation.
Advanced AI engine: The core of any touchless planning system is an AI engine equipped with sophisticated machine learning algorithms. These algorithms must be able to handle complex data sets, identify patterns, deal with outliers, and learn over time to improve forecast accuracy. They should be generalized enough so that they can be applied across different dimensions (such as for different products or markets) while also being mature and robust enough to consider various demand drivers at different levels of granularity. The AI engine's effectiveness directly impacts the system's overall forecasting performance.
High-accuracy results: The output of a touchless planning process should be a high-quality forecast that is accurate (low average forecast errors), unbiased (does not over- or under-forecast trends), reliable (avoids big outliers), and stable (low variations from one forecast to the next). Forecasts must accurately reflect future demand with minimal deviations, maintain consistency even in unusual circumstances, and avoid overreaction to minor data variations.
Explainable forecast: Explainable forecasts are essential for the successful adoption of touchless planning systems. In this context, explainability refers to a system’s ability to offer clear and understandable justifications for how it reached a particular decision or forecast (for example, high demand due to planned price reduction or social media activity). Easy-to-understand explanations help to increase a planner’s trust in a forecast, especially if the forecast is unusually high or low. It is also helpful to be able to quantify the system’s level of confidence in the forecast. All of this information should be presented in an intuitive manner through the user interface, allowing planners to easily understand and trust the system's outputs.
Feedback mechanisms: Touchless planning systems need to provide helpfulfeedback to human planners to stimulate their learning. If planners decide to adjust system-generated output, the system needs to provide detailed insights into how those adjustments affected forecasting accuracy and what the forecast value add (FVA) of the enrichments were. Without feedback, planners cannot learn which of their forecast enrichments or adjustments added value and which did not. In addition, systems should provide suggestions, such as identifying stock-keeping units (SKUs) suitable for touchless forecasting. This continuous loop of feedback and system guidance is essential for identifying improvement opportunities and boosting adoption.
Ultimately, all the above elements work in concert not only to improve the efficiency of the planning process and the planners themselves but also to foster trust, ensuring the successful adoption of touchless planning systems.
Business challenges
To gain deeper insights into the business challenges associated with transitioning to touchless planning, we conducted a series of interviews with supply chain leaders from a large number of companies. These companies are at different stages of implementation and span a range of industries. The top challenges mentioned include:
Knowledge gap. All too often, analytics and AI-driven decision-making operate on principles far from planners’ conventional expertise (for example, analyzing historical data, reconciling numbers across systems, or using traditional time series analysis). Due to that lack of understanding and training, many planners revert to manual enrichments instead of fully leveraging touchless planning.
Misalignment with reality. Part of the reason some touchless planning implementations fail to live up to expectations is that planners make adjustments to the forecasts that are not necessary. For example, a global beverage manufacturer reported that in some markets, when planners made significant adjustments to the forecast generated by the planning system, more than 90% of those adjustments failed to improve forecasting accuracy.
Forecast adjustments often stem from overoptimism about the success of a product launch or promotion. One interview partner remarked, “In my world, many forecast adjustments are full of dreams but not full of accuracy.” This comment highlights the discrepancy between wish and reality. Often, planners also inflate forecasts to increase inventories instead of adjusting safety-stock parameters in their planning systems.
In other cases, planners overcompensate for past events without recognizing the system's true capabilities. For instance, planners might adjust forecasts for high-demand events like Black Friday, not realizing the AI has already incorporated the promotional plans into its calculations. Or if past forecasts for the holiday season have been too low, planners might inflate future forecasts without understanding that the system had already learned from previous inaccuracies. These unnecessary adjustments show how planners lack understanding and trust in the system's logic and capabilities.
Shift in control. Planners often struggle with the fundamental mindset change required to transition to a touchless planning environment. Some planners may be resistant to no longer having a direct hand in the forecast. They may feel a loss of control and authority as systems become more autonomous. Yet at the same time, they still are accountable for the forecasting outcomes. This paradoxical situation is summed up by a manager at a North American food distributor as, "We're gonna get the blame whether it's a touchless environment or whether we did it ourselves."
Lack of regular, structured feedback. If a system does not have a good feedback mechanism, planners may feel uncertain about where to focus their efforts. In many settings, there is (not yet) a process in place that provides planners with the right data and feedback to build their trust in the system. As a result, planners may spend unnecessary time and effort on adjusting the outputs. Information on forecast value added (Did I improve the forecast?), clarity on demand drivers (Which factors are most relevant for this product?), and explainability of models (Why is the forecast that low?) are crucial.
Resistance to change. Organizations may be resistant to adopting new ways of planning. They may be concerned about the system’s accuracy and ability to understand changing market dynamics or its ability to react swiftly to unexpected changes. Despite all the advances made in touchless planning, some forecasts may inevitably err, potentially exacerbating resistance. This potentiality underscores the importance of alerts and guardrails to identify and catch these cases in advance. Building an understanding in the system's capabilities is critical toward gaining buy-in from planners.
Success factors
Based on our interviews, we also identified five success factors that companies need to consider when embarking on the journey to touchless planning. Getting things done right the first time is crucial for adoption. As one interview partner indicated, “It's very difficult to build trust if your initial implementation is not good. It's the first impression that lasts so long that it's just very hard to get away [from it].”
1.Standardize processes and conduct a “post-game” analysis: To ensure the success of touchless planning, it's critical to provide a common foundation for all the planning units involved. This foundation should include standardized processes and consistent reporting of key metrics such as forecasting accuracy, share of touchless volume, and forecasting value add. By having one common basis, companies can compare performance and track improvements more efficiently.
Furthermore, regular audits of forecasting performance can provide insights into potential improvement areas and ensure that the planning processes continuously evolve. Conducting post-game analyses using historical data can showcase the touchless system's performance, further convincing stakeholders of its value.
2. Establish training and a data-driven culture: Training for planners is indispensable when maximizing the capabilities of any planning software. This training should include how to identify demand drivers and understand data feeds. For example, planners should be trained to gather and input promotional information directly into the system rather than manually adjusting the forecasts for promotions.
In tandem with focusing on training, companies need to create a data-driven culture that emphasizes the transition "from qualitative to quantitative," as it was called by one interview partner. A quantitative culture will equip planners with the skills and mindset needed to leverage and curate data effectively.
3. Use transparency to create trust: Organizations can build and reinforce trust in AI-based planning systems by providing clear and timely feedback on the value add of enrichments made to the system-generated forecast. As one interviewee emphasized, “Seeing is believing.” Figure 3 shows an example of the type of feedback that can be provided to planners. The chart shows the forecast value add for a series of forecasting combinations (for example, the forecast for an item at a specific location for a certain time period). According to the chart, there are three zones: “touchless zone” where planners’ forecast adjustments are making results worse; “mixed zone,” where planners’ changes only provide a small value add; and “touch zone,” where planners’ changes are helpful, and their knowledge should be systematized.
FIGURE 3: Example of feedback on forecast value add (FVA)
Trust can also be fostered by providing transparency about how the AI came to its forecast. Understanding why a forecast is set to a certain value and not another, based on key demand drivers, helps to demystify the AI's decision-making process and to increase planners’ willingness to go along with it.
Planners also need tools that help them to decide when and where to “touch” or modify a forecast. This evaluation could be based factors such as historical accuracy, the commercial relevance of the forecast, or whether the discrepancies are already mitigated by safety stock.
4. Plan rollout strategically: Companies should initiate the touchless planning journey with a portfolio of low-risk decisions and gradually build confidence in the system’s capabilities. By starting small, companies can illustrate the system's effectiveness through real-world successes. It is often helpful to start by using touchless planning for items with high forecasting accuracy, lower volatility, or limited value at risk. From there, companies can follow different paths based on their maturity and risk attitude (see Figure 4).
FIGURE 4: Two paths to touchless planning
Organizations that aim to carefully understand the benefits of going touchless can start with AX items (indicating low volatility and high sales) that are well known to planners, who can then accompany the transition. Organizations that are less hands-on can observe results for less critical CX items (low volatility and low sales), before moving to either more volatile or more business-critical products.
5. Ensure leadership commitment: Finally, securing leadership sponsorship is essential for the successful implementation of touchless planning, ideally with the chief supply chain officer getting involved personally. A top-down approach—where leaders understand the key concepts, are data-driven, and are actively involved and supportive—sets a positive precedent. It's important to follow a blueprint that prioritizes process, people, and technology, in that order. Establishing a core team and an extended team, including commercial leaders, ensures comprehensive support across the organization.
Great potential
The potential for touchless planning within companies is immense, offering both enhanced forecast accuracy through advanced AI engines and improved planner efficiency by avoiding unnecessary interventions and overrides. However, planners must adapt to the new process that places more value on their contribution to inputs to the system and reduces manual output adjustments at the moment of truth. In touchless planning, planners need to trust the system to make decisions on their behalf. This requires embracing a loss of direct control and decision-making power as part of a more efficient and accurate data-driven forecasting process.
Keeping this in mind, it is crucial to create a clear vision for planners' future roles, emphasizing a shift from focusing primarily on overriding or adjusting the forecast to richer activities that involve more scenario planning and interactions with other functions, such as marketing and commercial teams. Consequently, planners' roles will encompass more value-adding tasks, such as understanding and providing insights into larger market trends, promotional levers, and customer behavior as well as curating the data fed into the system. Despite the shift towards automation, the need for human judgment remains, as not all decisions can be handled in touchless processes. Even as a company’s touchless planning process matures, we believe that there will still be situations that require traditional planning methods (see Figure 5). Ultimately, this ensures a promising outlook for planners’ future with no redundancies but richer job profiles.
FIGURE 5: Typical evolution of traditional to touchless planning
