1 of 9

BISG Supply Chain Lunch & Learn

March 5, 2024

Forecasting for the Rest of Us

Gary Jones – Partner, River Rock Advisors

Insights and Analytics for Business Decisions

2 of 9

In Prior Lunch & Learn Sessions We Discussed a Number of Concepts to Help Companies Achieve a Balance of Service, Cost and Capital Goals

All items are not the same, and therefore should have unique solutions
A number of item-level statistical forecasts can be “blended” to achieve a cross-functional, collaborative estimate
Statistical inventory techniques incorporate unique attributes of demand and supply variability
Detailed analysis of the implications of over- and under-ordering can help optimize replenishment quantities

Statistical
Sales
Customer
Finance

While this is true, not all companies have the data or resources available to implement such solutions

3 of 9

Alignment on Terminology and Expectations

According to Dictionary.com

Forecast (noun)

A prediction
A conjecture as to something in the future

Conjecture (noun)

The formation or expression of an opinion or theory without sufficient evidence for proof
An opinion or theory so formed or expressed; guess; speculation

According to Gary

Forecast (dream)

Estimation of the future, often given the illusion of accuracy with the use of decimals
“Lucky or lousy”…but we typically don’t know which it will be

But even with these disclaimers, we’re now going to discuss some potential forecasting tools!

4 of 9

“Forecasting for the Rest of Us”�Four “Simple” Steps to Generate Forecasts

Estimate a Total Quantity Over Time

Top-Down
Bottom-Up

Distribute the Estimated Quantity Over Future Time Periods

Profile-Based Allocation
Statistical Modeling

Apply Supplemental “Demand Intelligence”

Incremental Quantities Over Alternate Profiles
Manual Overrides
Multi-DFU Adjustments

Measure Results to Improve the Process

Maintain “Snapshots” by Forecasting Period, by Forecast Source
Establish & Calculate Key Forecast Quality Metrics…Over Different Lead Times
Utilize Forecast Performance and Trends to Initiate Process Improvement

DFU = Demand Forecasting Unit

Level of detail at which a forecast is established and adjusted
Examples:

National demand for an ISBN
Demand for an ISBN by State
Demand for an ISBN by State by Customer
Demand for an ISBN by State, by Customer, by Shipping Warehouse

5 of 9

“Forecasting for the Rest of Us”�1. Estimate a Total Quantity Over Time

Top-Down

Volume estimates are established at an aggregate level then distributed to individual Items through a hierarchy of DFUs

How the Total Quantities May Be Established

Prior year volumes – adjusted based on demand intelligence
Similar demand activity – item, customer, sales channel, etc.
Demand estimates – product development, sales, customers
Historical demand analysis – average and trend

Bottom-Up

Volume estimates are established at the individual Item level, with aggregations to higher levels of the hierarch of DFUs for validation

6 of 9

“Forecasting for the Rest of Us”�2. Distribute the Estimated Quantity Over Future Time Periods

Profile-Based Allocation

Establish a Library of Demand Profiles (% average demand by week)

Created by analyzing historical demand of items with “similar characteristics”

Apply the Appropriate Profile to the Estimated Quantity to Generate the Forecast

Statistical Modeling

Leverage Built-In Excel Functions to Generate Forecasts from Historical Demand

FORECAST.ETS – predicts future demand from historical demand based on Exponential Smoothing techniques

Provides options to enable potential seasonality detection

TREND – fits a straight line to historical demand to estimate a trend in volume (positive or negative)
Forecast Sheet – provides a number of features and options to enable Excel to generate a statistical forecast, based on historical demand

Function within in the “Data” tab

This is an advanced topic for another discussion

7 of 9

“Forecasting for the Rest of Us”�3. Apply Supplemental “Demand Intelligence”

Incremental Quantities Over Alternate Profiles

Establish a Library of Event Profiles (% event volume by week)

Created by analyzing historical Event impacts on demand
Estimated by Event planners

Estimate the total volume impact of the event
Apply the Event Profile to the estimated volume

Manual Overrides

When specific demand intelligence (or intuition) exists, adjustments to the forecast can be made manually

Expectation of specific customer orders
Competitive activity impacting demand
“Demand shaping” activities

Multi-DFU Adjustments

CAUTION: This is where “simple” forecasting can spin out of control
This is often utilized when a known event will impact a number of DFUs

Demand in the Northeast will decline by 10%
Demand to Retailers will increase by 7% in the Fall
Demand for Activity Books will decline by 3% in the Summer but increase by 5% in the Winter
Demand for books with a “Q” in the title will increase 250%

The challenge arises from DFUs that are impacted by a number of adjustments

Sequence of adjustments matter
Intended outcomes may not always be additive

8 of 9

“Forecasting for the Rest of Us”�4. Measure Results to Improve the Process

Maintain “Snapshots by Forecasting Period, by Forecast Source
Establish & Calculate Key Forecast Quality Metrics…Over Different Lead Times

Forecast Error – average percent error of the forecast
Forecast Bias – tendency of a forecast to be consistently high or low
Forecast Stability – degree to which the forecast changes from period to period

Utilize Forecast Performance and Trends to Initiate Process Improvement

Determine potential Quantity changes
Determine potential Profile changes
Evaluate the utilization of “Demand Intelligence”