Demand planning and inventory planning in a project-driven organization - the story of Ahrend
Case summary
Before this collaboration, Ahrend did not use a structured demand forecast to drive planning or inventory decisions.
Demand signals existed, but they were fragmented across historical data, CRM information, and planner judgment. None of them were strong enough on theirown to serve as a reliable reference.
We started by introducing a structured demand baseline at product line level. At that level, forecast accuracy improved by 7–34% compared to a basic statistical forecast on Ahrend’s most important product lines. Sales input was then added at the same level, mainly by pulling data directly from CRM. That input helped trigger conversations between sales and supply chain, but did not consistently improve the forecast itself on top of the CRM signals.
Only after that did we move the forecast down to lower levels to support operational and inventory decisions. This step required much more iteration, validation, and planner involvement, but it is also what unlocked the real value for inventory planning.
While the initial focus was on demand planning, the current phase is centered on using these detailed forecasts for inventory projections and execution. With better visibility across demand and supply, Ahrend is now able to steer inventory decisions with potential savings in the order of several million euros.
Company context
Ahrend operates in a project-driven manufacturing environment with a high degree of product configuration. Demand is driven by projects, framework agreements, and individual customer orders, each with different lead times, bill-of-material structures, and sourcing constraints.
In this kind of setup, demand planning is not about predicting exact orders. It is about creating demand signals that are stable enough to plan on, while accepting that a large part of demand will always remain uncertain.
The people driving the change
The work was led from within Ahrend by people who knew where the problems were and were willing to go into the details.
The project started under the leadership of Olivier Van Hoef, Senior Supply Chain Manager at Ahrend. With extensive experience in the company, Olivier’s objective was clear: bring more structure into demand planning, improve communication between sales and supply chain, and reduce structural overstocks.
As the project kicked off, Stephan Van Oorschot joined the core team. Stephan brought a lot of hands-on planning experience, including prior work with Slimstock and deep Excel-based analysis. His role became especially important for validation, collaboration with sales, and for the correct setup once the forecast moved beyond high-level views and had to work at the level where purchasing and inventory decisions are actually made.
Without their input and persistence, this would not have worked in this environment.
1. The starting point: fragmented signals and reactive planning
Before the project, demand planning relied on confirmed sales orders from ERP, pipeline information from CRM, and planner judgment. Each of these inputs made sense in isolation, but together they did not form a clear planning reference.
Forecasts were often either too detailed, at project or configuration level where patterns were unstable, or too aggregated to support real decisions. Inventory decisions followed the same pattern. Materials were ordered once projects were confirmed, or buffers were added manually to avoid shortages.
Sales opportunities were not always visible to supply chain at the right moment. When they did become visible, quantities were often adjusted late, which pushed inventory levels higher. The issue was not effort or intent, but the lack of a structured way to interpret demand signals.
2. Starting at product line level: building a baseline that works
The first deliberate choice was to start at product line level. At that level, demand patterns were stable enough to measure forecast quality and compare different approaches.
A structured demand baseline was introduced and benchmarked against Ahrend’s existing statistical forecast. Even without sales input or CRM signals, this baseline delivered 7–34% better forecast accuracy on the most important product lines.
This was an important checkpoint. It showed that the biggest improvement came from structure and level choice, not from adding more input. It also created trust before involving more stakeholders.
3. Sales input at product line level: useful conversations, mixed impact
Once the baseline was in place, sales input was added at the same product line level. This input came largely from pulling opportunities directly from CRM, without asking sales to create forecasts themselves.
In some cases, this improved the forecast. In other cases, it didn’t. The added value was not so much in the numbers, but in the conversations it triggered. Supply chain and sales started discussing upcoming projects earlier and more explicitly.
At this stage, sales input did not fundamentally change forecast accuracy, but it did improve alignment and visibility. That step was needed before going deeper.
4. Moving to lower levels: making the forecast operational
The real shift happened when the forecast was pushed down to lower levels to support operational and inventory decisions.
This is where things became harder. Patterns were weaker, data issues surfaced, and assumptions that worked at product line level no longer held. This step required detailed planner involvement.
Stephan spent significant time reviewing forecast behavior at lower levels, challenging aggregation logic, adjusting hierarchies, and refining time buckets. Several iterations were needed before the forecasts were consistent and stable enough to be used for purchasing and inventory planning.
This phase was not about squeezing out a few extra percentage points of accuracy. It was about making the forecast usable in practice.
5. Translating demand into inventory decisions
Once the demand signal worked at lower levels, inventory planning could finally move beyond reactive behavior.
Inventory targets were aligned with the demand structure, while purchasing decisions relied on the lowest-level forecasts, taking lead times, minimum order quantities, and sourcing constraints into account.
For the first time, inventory discussions were based on a shared and detailed view of future demand, rather than assumptions or safety buffers added late in the process.
6. Iteration, data work, and integration
Forecasts and inventory positions were reviewed repeatedly. Discussions focused on which CRM opportunities were realistic enough to include, how far ahead sales input should influence planning, how to deal with cancelled or delayed projects, and whether certain components should be treated differently.
Data alignment took time as well. Master data, historical demand, and product mappings had to be cleaned up, and IT availability meant that integrations were built step by step rather than upfront.
This progress came from detailed feedback by the planners. The setup evolved because they kept pushing on edge cases and inconsistencies until the outputs matched operational reality.
7. Results and current focus
After stabilization, demand planning became less reactive. Forecasts no longer shifted every time sales updated an opportunity, and inventory decisions could be made earlier and with clearer trade-offs.
More importantly, the detailed demand signal became the foundation for inventory projections.
As the focus shifts toward execution, these projections are now being used to quantify inventory impact and steer decisions that carry multi-million euro consequences.
The practical change is simple but important: Ahrend now knows earlier where it needs to prepare inventory, and where it does not need to build large buffers.
