The future of S&OP: How AI is revolutionising supply chain planning for paper & packaging Jari Kaukiainen, Advisor Integrated Business Planning and SCP, I Plan

PAPERTECHNOLOGYINTERNATIONAL
Transitioning from Manual to AI-Driven Forecasting
Many companies still rely on Excel or ERP-generated
static forecasts. Moving to AIdriven planning requires cultural and
capability shifts:
• Mindset Change: Planners shift from “owning the
number” to “guiding the system”. Instead of manually
manipulating forecasts, they now oversee automated
predictions and validate them with business context.
• Onboarding Support: At I-Plan, we provide structured
onboarding programs and hands-on training to help
teams adapt. Visual analytics and intuitive dashboards
make adoption smooth.
• Confidence in the System: Early wins build trust. For
example, one planner at an I-Plan customer noted,
“Before I-Plan, we were constantly fire fighting. Now we
can finally think ahead to grow the forecast value-add.”
The Human Element: Why Planners Still Matter
AI doesn’t replace planners— the use of AI rather empowers them.
Many limitations still exist:
• AI can’t interpret politics, contracts, or intuition
• Poor data leads to poor outputs
• Explainability remains a challenge in some models
Human oversight remains essential. Planners are now
strategic enablers, guiding the AI, injecting context, and ensuring
cross-functional buy-in.
A client planner put it best: “With I-Plan, we’re not just
planning better—we’re finally having the right conversations at the
right time.”
Forecasting Tailored to Every SKU: The Power of Expert
Selection
I-Plan’s “Expert Selection” tool addresses the reality that no
one-size-fits-all approach works for all SKUs. It dynamically selects
from 14 statistical methods, including models like:
• Exponential Smoothing
• Box-Jenkins (ARIMA)
• Moving Average
• Croston’s Method
• Delphi Method
Each SKU is analysed individually, ensuring the most
suitable model is used and updated continuously. For companies
managing thousands of SKUs, this level of automation delivers
speed, accuracy, and confidence.
Improving Forecast Accuracy with MAE, MAPE and Forecast
Value Add (FVA)
Forecast accuracy is a critical performance metric in supply
chains with high variability and long lead times—like in the paper
and packaging industry. Among the many KPIs, three stand out for
their practical value: Mean Absolute Error (MAE), Mean Absolute
Percentage Error (MAPE), and Forecast Value Add (FVA).
Forecast MAE measures the average absolute difference
between forecasted and actual values in real units (e.g., tons). It’s
easy to interpret and unbiased toward over- or under-forecasting
but is not scale-independent—making comparisons across products
challenging.
MAPE, by contrast, expresses forecast error as a
percentage, making it ideal for comparing performance across
SKUs or business units. In I-Plan’s Expert Selection tool, multiple
statistical methods are applied at the SKU level, and the model with
the lowest MAPE is selected—often resulting in forecast accuracy
improvements exceeding 20% within weeks.
While MAE and MAPE assess statistical accuracy, FVA
asks: Did human adjustments improve the forecast or make it
worse? It compares different versions of the forecast:
• Naïve forecast (e.g., last period’s actuals)
• System-generated forecast (statistical or ML-based)
• Planner-adjusted forecast
• Final consensus forecast (e.g., post-S&OP)
FVA helps identify where human input adds value—and
where it introduces bias. For example, if a sales override increases
MAPE from 20% to 32%, FVA highlights this as a negative
contribution, showing the system forecast would have been better
left untouched.
78

Paperturn flipbook viewer