In the previous article, I discussed three lessons learned and observations that emerged during a retail planning transformation. The first was that more detail does not always lead to better planning. The second was that not every product deserves the same amount of planning effort. The third was that some of the most valuable planning decisions occur when planners choose not to intervene.
These observations ultimately led to a broader question. If planners should spend less time maintaining forecasts and more time focusing on meaningful exceptions, what should the planning process look like instead?
There is no universal answer to that question. Every organization operates with different products, planning cycles, and business requirements. However, one of the lessons learned during this transformation was that creating a scalable planning process required more than a change in forecast methodology. It required changes to the way forecasts were constructed, how planning decisions were represented, and how the IBM Planning Analytics solution itself was developed and maintained.
While forecasting curves, layered planning structures, IBM Planning Analytics implementation, and AI-assisted development may appear to be independent topics, they were ultimately introduced to support a common objective: helping planners spend less time maintaining forecasts and more time making informed business decisions.
Building forecasts from observed behavior
One of the challenges faced by many planning organizations is establishing consistent forecast starting points across a large assortment. Even when planners are working with similar products, differences in planning approach, individual judgment, and historical processes can result in forecasts being constructed differently across categories, locations, or business units.
As part of the transformation, the team explored the use of analytical curves derived from actual product behavior. Rather than expecting planners to create every forecast from scratch, these curves provided a structured starting point based on observed demand patterns.
This approach was particularly valuable because many products share common demand characteristics. Some products experience a rapid increase in demand after launch before stabilizing. Others exhibit seasonal patterns that repeat year after year. While no two products behave identically, many follow broadly similar trajectories that can be represented and reused.
Forecasting curves were not intended to produce a final answer or remove planners from the process. Instead, they reduced the need to rebuild common demand assumptions each planning cycle, allowing planners to focus their judgment on the products and circumstances that required a different approach. As assortment complexity increased, this consistency became increasingly valuable because it reduced the amount of effort required to establish a baseline forecast and allowed planner attention to be directed toward products where additional insight was required.
Separating forecasts from planning decisions
As forecasting approaches evolved, another challenge became apparent. Many planning environments combine multiple decisions into a single forecast value, making it difficult to understand what changed, why it changed, and who made the change.
Promotional impacts, top-down guidance, local market adjustments, planner overrides, and baseline forecasts are often blended together until the final number is presented without clear visibility into the individual contributors.
During the transformation, the team adopted a layered planning approach that separated these decisions into independent components. Baseline forecasts, promotional impacts, top-down inputs, and planner adjustments could each be represented independently while still contributing to the final forecast.
The practical benefit of this approach was transparency. Rather than debating a single forecast value, stakeholders could understand the factors influencing the result and examine individual assumptions independently. Discussions became more focused because participants could challenge specific drivers rather than attempting to explain changes embedded within an aggregated number.
One of the more interesting lessons learned was that many planning disagreements were not actually disagreements about the forecast itself. More often, they were disagreements about assumptions that had become difficult to identify once they were combined into a single value. By separating planning inputs into independent and auditable layers, the organization created a planning process that was easier to understand, explain, and maintain.
Modernizing the planning platform
The planning process itself was only one part of the transformation. The retailer’s existing merchandise planning solution had evolved internally over time and relied heavily on spreadsheets and custom-developed processes. While that approach provided flexibility and addressed the organization’s immediate planning needs, it became increasingly difficult to maintain as the assortment, planning requirements, and number of users continued to grow.
The implementation of IBM Planning Analytics as a Service on AWS introduced a new planning foundation and provided an opportunity to reconsider how forecasts were created, how planning decisions were represented, and how planners interacted with the overall process. Rather than attempting to recreate the existing solution in a different technology, the team used the transformation to evaluate which aspects of the prior process should be retained, which should be standardized, and which could be redesigned to better support the organization’s future planning needs.
This distinction was particularly important because internally developed solutions often reflect years of valid business decisions, specialized requirements, and accumulated knowledge. Moving to a new platform does not eliminate the value of that experience, but it does provide an opportunity to determine whether processes that developed incrementally still represent the most effective way of working. In this case, forecasting curves, layered planning structures, and exception-based workflows helped translate important business requirements into a more integrated and scalable planning model.
One of the lessons learned was that modernization initiatives are most successful when process and platform design evolve together. Simply reproducing an existing solution can carry forward much of the complexity and technical debt that prompted the transformation, while redesigning the process without understanding the business requirements that shaped the original solution can create unnecessary disruption. The team therefore needed to balance modernization with an understanding of why the previous processes existed and what planners still needed from the new platform.
IBM Planning Analytics as a Service on AWS served as the foundation for that broader transformation. Its role was not simply to replace the technology supporting the prior process, but to create a planning environment that performed effectively, met the organization’s functional requirements, and was practical for planners to use and maintain. This allowed the organization to preserve the business knowledge embedded in its existing practices while reconsidering how planner effort could be directed across the assortment.
AI-assisted IBM Planning Analytics development
The implementation of IBM Planning Analytics also provided an opportunity to incorporate AI-assisted development tools throughout the design and delivery of the new platform.
Much of the current discussion around AI focuses on forecasting and decision-making. While those topics are important, one of the most immediate opportunities observed during this project was within the development lifecycle itself.
IBM Planning Analytics implementations involve a significant amount of iterative work, including model development, documentation, testing, process design, and supporting technical artifacts. Many of these activities follow predictable patterns and benefit from consistency. Throughout the implementation, the team leveraged AI-assisted development tools and MCP-based capabilities to accelerate certain development activities while allowing developers to spend more time focusing on architecture, business requirements, and solution design.
The value did not come from replacing expertise. Rather, it came from reducing the effort associated with repeatable tasks and improving consistency across deliverables. Equally important, the project reinforced the need for experienced developers and planners to validate AI-generated outputs. While AI could accelerate development activities and generate initial artifacts, solution quality still depended on human review, testing, and validation to ensure the results aligned with business requirements and operated as intended.
One of the lessons learned was that AI can significantly improve productivity, but it does not eliminate the responsibility for design reviews, testing, quality assurance or change management. Those activities remained critical throughout the implementation and continued to rely on experienced practitioners applying business knowledge, planning experience, and architectural judgment.
As organizations continue to evaluate AI within IBM Planning Analytics implementations, the distinction between accelerating work and replacing expertise will become increasingly important. The most effective use cases may not be those that attempt to replace planners or developers, but those that allow them to focus more of their time on higher-value activities.
Bringing it together
Although forecasting curves, layered planning structures, IBM Planning Analytics modernization, and AI-assisted development address different challenges, they were all introduced to support the same objective.
Forecasting curves created consistent planning foundations. Layered planning structures improved transparency and made planning decisions easier to understand. IBM Planning Analytics provided the foundation for replacing an internally developed planning solution with a more integrated and scalable planning environment. AI-assisted development accelerated implementation and helped improve consistency across the solution, while human review and testing remained essential for validating outputs and ensuring alignment with business requirements.
Together, these changes helped create a planning environment that reduced routine maintenance while preserving the role of planner judgment. More importantly, they supported the broader planning philosophy described in the first article: organizations create value not by reviewing every forecast, but by helping planners focus their attention on the decisions where their expertise can have the greatest impact.
Conclusion
Planning complexity is unlikely to decrease in the future. Assortments will continue to grow, planning cycles will continue to accelerate, and expectations for responsiveness will continue to increase. While modern planning platforms can generate forecasts at scale, organizations still need practical ways to manage the volume of decisions created by that scale.
The approaches discussed in this article represent one organization’s response to that challenge. Forecasting curves, layered planning structures, IBM Planning Analytics modernization, and AI-assisted development were not objectives in themselves. They were components of a broader effort to create a planning process that could scale while allowing planners to focus on meaningful exceptions.
Ultimately, the goal was never to create more forecasts. The goal was to ensure that planners could spend more of their time making decisions and less of their time maintaining them.
If this topic is of interest to you
If this topic is of interest to you, I encourage you to join me at IBM TechXchange 2026 for my session, “Curves Do the Work: Planners Handle Exceptions,” where I’ll discuss how these concepts were applied during a retail planning transformation and share some of the lessons learned along the way. We’ll explore how IBM Planning Analytics as a Service on AWS, forecasting curves, layered planning approaches, and AI-assisted development supported a more scalable planning process while helping planners focus on the decisions that matter most.
Kind regards,
Rose Morgan
Cubewise Practice Manager