The Planner's New Role in an AI-Forecasting World: Measuring Forecast Value Add

In practice: A client case study for forecast value-add 

With an AI-led forecasting solution, FVA is not a "nice to have" report you get to eventually. It's the foundation of the process that defines how planners and AI work together to achieve the highest possible forecast accuracy. It's part of e2open's standard solution offering, and where we spend most of our time with clients. 

A global industrial and technology company began using e2open's long-term Demand Sensing solution, with e2open creating the baseline forecast using AI. The company then ran that forecast through a consensus process, in which planners made adjustments based on their own judgment and on input from the sales team. 

Once FVA was in place, the data told a clear story: planner overrides were reducing forecast accuracy across nearly every tier. Expressed in mean absolute percentage error (MAPE) points, a negative FVA means an adjustment made accuracy worse; a positive FVA means it made it better. 

On the highest-volume products (Tier 1), manual adjustments were essentially a wash. The further planners moved into lower-volume, longer-tail items, the more damage those adjustments did; up to a -55% worse on the lowest-volume, intermittent-demand items. When the best outcome achieved was an FVA of 0%, it meant that the time required by planners to make adjustments had little benefit and often degraded the forecast accuracy. 

Viewing the data by product lifecycle told a similar story. New and ramping products, for which historical data didn’t exist, experienced significant degradation in forecast accuracy. Mature products, which make up a majority of the portfolio, saw a much smaller decline, at -3%. 

The one bright spot was with phase-out items. Planner adjustments improved forecast accuracy by 6% over the engine-generated forecast. This is often a strong spot for planner input because they can apply direct customer visibility into last-time buys and final orders that the AI couldn't see.  

That is the template for when a planner should step in: not everywhere, but precisely where they know something the system lacks. 

That said, these are category-level metrics. Within every segment, there were individual products where planner adjustments improved the forecast. The numbers tell us where to focus, not that people never add value to a category or that the machine is always right.  

The best outcome comes from the right blend of the two: the AI handling the patterns it reads better than any person can, and planners stepping in where they hold information the AI can't see. 

Perhaps counterintuitively, the company took these results as a win. The tool was driving a great outcome. The conversation then shifted from trying to optimize human judgment across every forecast to a more pointed question: when and where should people actually intervene?  

A new generation of the e2open Demand Sensing solution 

At e2open, we’re very excited about our next generation of Demand Sensing. Recently, e2open rebuilt the product from the ground up to support the growing size and complexity of global supply chains. The new architecture handles greater scale, gives planners more control and configurability, and makes it easier to adopt new AI advances as they emerge. 

A key example is how we reimagined the user experience based on how value typically unfolds when adopting new demand forecasting software. In year one after implementing a new technology and process, companies often see substantial gains in forecast accuracy. But sustaining double-digit improvements of 10%, 20%, or 30% year-over-year is extremely challenging. The low-hanging fruit disappears, and continuous improvement requires a fundamental shift in how planners work. The goal changes: instead of chasing one big 10% win, you start looking for ten 1% wins. 

This is one problem our next-generation Demand Sensing product is built to solve. A planner can start with a hypothesis, for instance that feeding additional data into the forecast engine would improve accuracy. They can then simulate it in a scenario to back-test whether it actually improves forecast accuracy. The system tells you whether the idea works, where, and by how much. If it helps, you know it's worth setting up that data pipeline going forward; if it doesn't, you simply move on to the next idea.  

The result is a built-in cycle of hypothesis, measurement, and continuous improvement, so value doesn't plateau after year one. It compounds. Planners get better forecast accuracy, clearer workflows for collaborating with AI, and smarter use of IT resources, since you only build the data pipelines that actually move the needle. 

Bottom line: A new role for planners 

So, what does the planner do now? Less manual work, like picking the right algorithm and deciding which inputs matter and by how much, which is all automated by the tool.  

Instead, they focus on the changes that measurably improve the business. They manage the quality of what goes into the engine, they guide the engine with business context, and they intervene only where they hold information the system cannot see. FVA is what makes this possible, turning the collaboration between people and machine into a clear, measurable map of where each one adds value. The forecast planners of the future are not judged on the forecasts they produce, but on the value they add. 

This is the kind of work the new generation of e2open Demand Sensing is designed to support. It gives planners the tools to find their next source of value, test it, measure it, and act on it. That is how the first-year gains turn into value that compounds over years.  

Focusing where planners are really adding value. Chasing the next 1%. That's the job now. 

Frequently Asked Questions 

What is Forecast Value Add in demand planning?  

Forecast Value Add, or FVA, is a way to measure whether each step in the forecasting process improves or hurts forecast accuracy. In an AI demand forecasting environment, it compares the engine-generated forecast with planner adjustments and actual demand, showing where human input adds measurable value and where the forecast engine performs better on its own. 

How does AI demand forecasting change the role of the demand planner? 

AI demand forecasting shifts the demand planner’s role from manually building forecasts to orchestrating better outcomes. Planners focus on data quality, business context, exception management, and Forecast Value Add analysis so they can intervene only where their knowledge improves the forecast. 

Why is Forecast Value Add important for forecast accuracy improvement? 

Forecast Value Add is important because it turns forecast accuracy improvement into a measurable, repeatable process. Instead of assuming every planner override is helpful, FVA shows which adjustments made the forecast better, which made it worse, and where planning teams should focus their time for the greatest impact. 

What is the difference between consensus forecasting and an AI-generated forecast?  

An AI-generated forecast is created by the forecast engine using historical data, real-time signals, machine learning techniques, and configured business parameters. Consensus forecasting typically starts with that baseline and incorporates human judgment and cross-functional input from teams such as sales, marketing, and supply chain planning. 

How can demand forecasting software help planners find the next source of value?  

Modern demand forecasting software can help planners test hypotheses, simulate new data inputs, measure forecast accuracy, and identify where small improvements can compound over time. This supports a continuous forecasting process in which planners focus on the changes that measurably improve business outcomes.