FORECASTING
101
IT’S ALL ABOUT
Ed White
CPIM CIRM CSCP CPF LSSBB
is the founder of Jade Trillium
Consulting. Over the course of his career he has worked in several
industries ranging from food and pharmaceutical to chemicals and
plastics. He has held positions at many different management levels
and been responsible for numerous projects in Supply Chain, cost
saving, education and value enhancement,
Ed is an active member of APICS as an instructor and Chapter Board
member. In addition Ed is a published author and sought after
speaker on various Supply Chain topics. He has spoken at several
international conferences for APICS, IBF & SAPICS.
Biography
FORECASTING IS LIKE BREATHING,
EVERYONE DOES IT AND YOU
How long will it take to get to work?
•
What time do you have to leave
•
Based on history, modified by weather conditions, current
road construction and traffic conditions
Will I run out of money before I run out of
month?
•
Decide whether to go see a movie with family
•
Based on outstanding bills, modified by current cash
situation and understanding of potential new bills
There are two types of forecasts:
FORMAL:
• Defined process and procedures
• Used by everyone in the defined group
• THIS IS WHAT YOU WANT!!!
INFORMAL:
• No defined process
• Everyone making their best guess as to what will be needed in their
own areas
Because we want to:
to keep our customers happy and
profitable
so we can stay happy and
profitable
to ensure material is available
for sale
at an acceptable customer
service level
while maintaining stock levels
at a financial level that is
acceptable to the company.
in the right place
at the right time
in the right quantity
at the right cost
Purpose of MRP is to recommend changes to
material stock levels that will;
Forecasting is the prelude to planning
Little planning can be done without some
form of estimation
Forecasts are at best an informed guess
Forecast is NOT the budget divided by 12
Forecast is related to the LE not the
Are almost always wrong
•
If they are right it’s a mistake
Are almost always wrong
•
If they are right it’s a mistake
Should include an estimate of error
Are more accurate for groups of products
Are more accurate for nearer periods of
There are two types of formal forecasts:
Quantitative
•
Calculated based on
someone’s past sales data
Qualitative
The problem with only using
a calculated forecast is that it
is like driving using only the
rear view mirror.
You’re OK until the first
curve in the road and
•
Efficient production sequencing which will lead to more
available capacity.
•
Less rework
•
Better run size distribution
•
Reduced Dead and Slow Inventory due to a more accurate
product mix in the inventories and greater inventory turnover
•
Improved product quality
•
Improved customer confidence
•
Cost containment through improved freight logistics
•
Optimized inventories
•
Inventory carrying cost on Dead and Slow
Moving Stock
•
Dead and Slow Write-Downs
•
Extra warehousing costs for ineffective
inventory
•
Lost production due to rush inserts
Short answer – everyone that has
information on future demand.
System can generate a calculated
forecast based on history
Sales must modify based on Cust.
knowledge
SALES & OPERATIONS PLANNING (S&OP) process
is used to arrive at a consensus forecast
– Only sales knows Cust. long
term plans.
Trend – General movement upwards or
downwards
Seasonal – repeating pattern over a period
of time
Bias – artificially introduced trend
Outlier – data point that is significantly
different than the norm
General business and economic conditions
Competitive factors
Market trends
•
such as changing demands in the marketplace
Firm’s own plans
•
pricing
•
product changes
Jan
100
Feb
90
Mar
110
Apr
105
May
95
Jun
90
July
90
Aug
120
Sept
?
Simplistic – Last month sales repeated
• This is the simplest form of formal
forecast available.
• Only useable if sales very steady
Since Aug sales was 120 units the forecast for Sept is 120
• To allow for planning lead time you could always
offset by one or more months. For example a one
month offset would use July sales of 90 units as the
forecast for Sept.
Jan
100
Feb
90
Mar
110
Apr
105
May
95
Jun
90
July
90
Aug
120
Sept
?
Simplistic – Last month sales repeated (Forecast =120)
Average – Average preceding 12 months sales
• Ignores seasonality and variability
In this example we only have 8 months of data so an
8 month average would be (100 + 90 + 110 + 105 +
95 + 90 + 90 + 120)/8 = 800/8 = 100
Jan
100
Feb
90
Mar
110
Apr
105
May
95
Jun
90
July
90
Aug
120
Sept
?
Simplistic – Last month sales repeated (Forecast =120)
Average – Average preceding 12 months sales (Forecast = 100)
Moving 3 month Average – Average only
previous 3 month sales
• Captures seasonality but not trends
Previous three months (Jun, July and Aug) give an
average of (90 + 90 + 120)/3 = 300/3 = 100
Jan
100
Feb
90
Mar
110
Apr
105
May
95
Jun
90
July
90
Aug
120
Sept
?
Simplistic – Last month sales repeated
(Forecast =120)
Average – Average preceding 12 months sales
(Forecast = 100)
Moving 3 month Average – Average only previous 3
month sales
(Forecast = 100)
Winters, Student T, Multiple Linear Regression, etc
• Potentially more accurate but requires
statistical background to set up and interpret
1. You want the formula and process that gives you
at least the lowest acceptable accuracy
2. This should be balanced against the cost of the
process. If you are spending more than you are
saving, find a simpler process.
3. Should always compare your accuracy against
the accuracy of simplistic formula. If no major
improvement, why use the more complicated
formula / methodology.
1. Are there any one time events like
shutdowns
planned by the customer?
2. Are there any anticipated increases in demand?
3. Are there any anticipated decreases in demand?
4. Is there any chance of an existing product / size
being replaced by a new combination?
5. Is there any chance of a change to customer /
location / etc?
• No more than 3 hours a month
• Critical customers first
• Critical materials second
• Low f/c accuracy combinations
before high accuracy combinations
• If you must guess - guess high
• Forecasts are always wrong
• We don’t care if they are wrong
• The important point is how wrong
are they!
By monitoring the forecast accuracy report we
can track which material / customer
combinations or other groups of forecasts
have the most opportunity for improvement.
REMEMBER: FORECASTS ARE MORE ACCURATE
• FOR LARGER GROUPS
•
Therefore more accurate at material level rather than at
material / Cust. Level
• IN THE NEAR FUTURE RATHER THAN THE DISTANT FUTURE
• 3 months out more accurate than 12 months out, but 12
months still important for tentative capacity planning
• FOR A CUMULATIVE TIME FRAME RATHER THAN A SINGLE
TIME FRAME
Formula: Forecast Accuracy (%) = x 100
1st Step:
Calculation of absolute
forecast error with respect to
forecast lowest level
ONLY
2nd Step:
Examine result for special
cases at the article level
• If Forecast = 0 and Actual
0, then Forecast Accuracy = 0%
• If Actual = 0 and Forecast = 0, then Forecast Accuracy = 100%
• If Forecast
0 and Actual = 0, then Forecast Accuracy = 0%
2.
1.
Sales - Sales- Forecast
Sales
Month
Jan
Forecast
from Dec
2000
1800
Actual
Sales
200
Sales - Forecast
3.
3rd Step:
Calculation of Forecast
Accuracy by using given
formula
Forecast Accuracy (%) = x 100) =
89,9%
1800 - 200
1800
4.
4th Step:
Aggregation according to
predefined reporting levels
• Aggregation of
SALES
and
absolute deviations
on all
aggregation levels; calculation of forecast accuracy on all
aggregation levels
• Consequence: Deviations grow on every aggregation level
• The important factor is not technique, it is process control
and management
• The S&OP process allows everyone to participate in
creating a consensus Demand & Supply plan
• Anyone can forecast but the right person with the right
information can significantly increase accuracy
• Small changes in forecast accuracy can translate to big $
improvement in profitability
FORECASTING IS JUST ANOTHER
COMMUNICATIONS PROCESS
Remember,
IT’S ALL ABOUT
For more information or if you have any
comments about this presentation please
contact me at:
[email protected]
Please remember to give your completed
feedback form to our room monitor.
Forecasting 101
It’s all about the customer
[email protected]
905-483-5984
http:jadetrilliumconsulting.com
Ed White@JadeTrillium
ca.linkedin.com/in/edwhitesupplychain/
Ed White CPIM CIRM CSCP CPF LSSBB
Value Catalyst
Jade Trillium Consulting
FORECAST PROCESSES
rounded Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total Last year sales 9,250 9,100 9,000 9,050 8,950 9,250 9,250 9,400 9,550 9,675 9,900 9,850 112,225
Simplistic
last yr + 10% Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total This year f/c 10,175 10,010 9,900 9,955 9,845 10,175 10,175 10,340 10,505 10,643 10,890 10,835 123,448
rounded Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total Last year sales 9,250 9,100 9,000 9,050 8,950 9,250 9,250 9,400 9,550 9,675 9,900 9,850 112,225
Adjusted
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total This year f/c 10,175 10,010 9,900 9,955 9,845 10,175 10,175 10,340 10,505 10,643 10,890 10,835 123,448
Sales Adjustment - 100 - 100 - 50 50 - - 100 200 250 250 400 1,100
Final f/c 10,075 9,910 9,900 10,005 9,895 10,175 10,175 10,440 10,705 10,893 11,140 11,235 124,548
rounded Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total Last year sales 9,250 9,100 9,000 9,050 8,950 9,250 9,250 9,400 9,550 9,675 9,900 9,850 112,225
Tiered
This year f/c Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total
Promotional - 100 - - 100 - 50 - - - - 100 400 400 850 Holiday - - - - 100 200 300 Seasonal 100 - 100 - 200 - 300 - 200 - 100 - 100 300 500 400 300 800 Internal 50 50 50 50 50 50 50 50 50 50 50 50 600 Service 100 100 100 100 100 100 100 100 100 100 100 100 1,200 Base 10,000 10,000 10,000 10,000 10,000 10,000 10,000 10,000 10,000 10,000 10,000 10,000 120,000