Maintaining credible supply chain plans through demand and forecasting uncertainty

With unprecedented levels of market and geopolitical uncertainty, supply chain planners are being asked to commit to numbers they cannot justify, built on inputs and models they no longer trust. 

And when the baseline is wrong, downstream decisions inherit and amplify the error. It shows up everywhere, from safety stock and capacity commitments to hiring and working capital.

Instead of getting to one right number, planners need to work from a unified ranged forecast that evolves with shifting signals and market conditions.

Unreliable markets create unreliable planning conditions

Unpredictability is affecting every corner of the retail industry, from supply chain reliability to lead time and tariffs. Both retail companies and manufacturers are feeling the pressure:

• 78% of manufacturers now name trade uncertainty as their top concern, and many companies have responded by front-loading inventory and rebuilding supply chain structures. 
• 68% of companies consider lead time variability to be the top supplier challenge, ahead of long supplier lead times at 58%. 
• 82% of companies report that new tariffs are affecting their supply chains, due to pull-forward buying ahead of tariff deadlines. 

The reality is last year's figures no longer say much about next quarter. To remain nimble during times of volatility, planning teams must embrace flexibility while ensuring decisions are driven by current data.

This article lays out three steps that help businesses do just that, even when it feels like concrete numbers are out of reach.

Step 1: Widen the point forecast with a range

Many organizations still produce a point forecast that they treat as the authority. Teams build their plans around it, commit budgets against it, and staff against it.

Introducing a range around that point forecast helps cover the variance one number alone would miss. For example, demand might have an 80% chance of landing between 600 and 800 units and a 20% chance of coming in lower. 

If you were to average that figure out to 700 to give your team something concrete to plan around, you’d be hiding the 20% chance of falling short.

The best planning tools will allow you to do all of the above, while maintaining explainability. In that same example as above, the system should tell you your plan has an 80% chance of landing between 600 and 800 units - and here’s why. 

A range also holds up better when the underlying data isn't perfect. For the 76% of treasurers citing poor data quality as a top challenge, that’s a reality that consistently makes planning harder. 

Pinpointing only one figure demands a level of precision flawed data cannot reliably deliver. A range, on the other hand, uses the imperfect data while still offering a reasonable projection that leadership can stand behind.

Step 2: Use AI to keep the forecast current between cycles

Traditional forecasting models retrain on a fixed cadence, often monthly or quarterly. Even with predictable market conditions, by the time a new cycle starts, the assumptions inside the model are already out of date. 

AI-driven forecasts work differently. They retrain continuously, incorporating new signals as they arrive instead of waiting for the next scheduled review. This continuous approach is already becoming standard practice, with adoption reaching 87% among leading organizations.

One of those organizations is Coca-Cola North America, where Sara Park's team built an AI-powered decision intelligence model with Pigment that learns from planners' fulfillment decisions over time. When an order arrives and inventory sits in the wrong location, the model presents planners with fulfillment options: pull stock from another location, break into the production schedule, or fulfill from a different warehouse. It calculates the cost and revenue impact of each option—a step planners previously had to work through manually before choosing one.

As the model learns which option planners tend to choose given the variables at play, it starts proactively surfacing the best options. Because its recommendations stay current with how conditions are actually changing, planners no longer have to manually work through every possibility before choosing one. They can act on the model’s recommendation and move on.

Step 3: Give every team the same number to plan against

A defensible, current forecast still won’t be effective if commercial, brand, and supply chain teams are each defending a different one. When each team forecasts in their own spreadsheets, an organization ends up with multiple, sometimes contrasting views of demand. There's no consistent number or range to plan against, each handoff introduces version drift, and planning turns into a reconciliation exercise instead of decision-making.

On the other hand, when the entire organization uses a unified model, forecasts become a shared artifact every function directly plans against. A change to any input updates the model everywhere in real time. Demand, supply, and financial views stay in sync without anyone rekeying numbers.

Once you’re all looking at the same numbers, it’s also important to be aligned on the risks and opportunities that might make them move. That way, have sufficient coverage and  be ready to react.

Inside one retailer's integrated demand planning process

A global fashion retailer operating across 27 countries built exactly this kind of unified process on Pigment. It replaced a fragmented setup where brand, commercial, and supply chain teams had each worked from isolated spreadsheets.

After standing up a dedicated demand planning team in 2023, one fashion retailer implemented Pigment for supply chain planning.

The team now consolidates commercial inputs from sales across regions and channels with product-focused inputs from brand teams into a single integrated planning model: their integrated demand planning (IDP) process. From there, they run base and drive models to test different assumptions and surface what would have to be true to hit each target. 

The same model also supports weekly inbound and processing plans against warehouse capacity, monthly buy plans and on-the-fly scenario tests. For instance, the model can simulate switching certain purchase orders from ocean to air freight when warehouse capacity opens up, weighing the cost trade-offs and surfacing a recommendation for procurement to act on.

The model’s real-time exception reporting flags risks such as delayed container departures before they become a problem, contributing to up to a 20% seasonal improvement in delay mitigation. 

As the fashion retailer’s planning team has gained confidence with its new model, it’s also grown increasingly self-sufficient in building its own tools. More than half of the company's planning applications are now built directly in Pigment by team members themselves, a sign of a broader shift toward empowered, data-driven decision-making.

Reliable forecasting is possible with a shift in mindset

Uncertainty likely won’t change, but the way planning teams approach forecasting can. 

A team that plans around a range, leverages a continuously updating model, and operates from one shared number spends less time redoing plans when reality shifts. 

That time goes toward more valuable work: deciding where to shift capacity, which risks to absorb, which scenario to prepare for next.