From Forecasting to Drink and how we could be more sociable with business

1
From Forecasting to Drink and how we could be
more sociable with business

• Peter Gormley, Business Development Manager,
  Gordon MacMillan, Promotional Analysis Manager,
• Scottish Courage Ltd.

2
Scottish Courage Brands Ltd.

• Part of Scottish Newcastle plc
• 26 domestic share, 30 core brands own label
• 250 SKUs, 130 new each year
• 200 staff, 800m turnover, over 60m profit
• Market - Interbrew, Coors, Carlsberg, A-Busch,
  Guinness
• 11.3 million barrels, underlying growth 4 per
  annum
• 70 of volume from 3 brewers
• 53,000 outlets, but 4 store groups (1700 stores)
  30
• 500 brands, but top 13 brands gt half of volume
• Take Home 31 of UK beer market USA - 70,
  Germany - 65, France - 61, Ireland - 10

3
Criticality of Forecasts

• Sales Operations Planning - total beer business
  - 2 yr.
• All aspects of planning - sales, marketing,
  finance, supply..
• Pricing and promotional activity - 60 sold on
  promotion
• Impacts on service, stock, waste, efficiency,
  profit
• On-trade stable, off-trade highly volatile
• Polarisation - grocers, wholesale, specialists,
  convenience..
• Price and promotional offers, BOGOFs,.
• In-store display and feature, events, weather,
  competitors..
• Promiscuous, elastic market
• Highly seasonal

4
Becks Bier Supply to Major Customer
11.49
11.49
12.49
12pk BOGOF
12.99
12.49
11.99
12.49
12.99
5
Forecast Process Evolution

• Output - forecast by customer by SKU by period -
  2 years
• Statistical forecast based on supply data
• Sales Marketing edit forecast at various
  horizons
• Assumptions captured in database
• Valuation of forecast
• Forecast review meetings and submission to group
  SOP
• Move to top down forecast managed by one function
• Information passed from Sales Marketing
• Price and promotion models used

6
Demand Factors
Demographics
Display
SCB Pricing
Stubbies
Promotional Effectiveness
ECR Initiative
Weather
Parallel Imports
Opening Hours
Economy
Duty
Events
Multibuy
Customer Performance
Health Conscience
Calais
Leisure Time
Competitive Pricing
On Pack Offers
Brand Loyalty
In Home Entertainment
Recreational Drugs
Legislation
7
Lancaster Regression Models

• Different levels of forecast
• Considered
• price, price differential, media spend,
  promotion, multibuy, display, feature,
  temperature, sunshine, seasonality, distribution,
  etc.
• Regression outperformed exponential smoothing
  model
• 10 MAPE vs. 15 for total beer
• 17 MAPE vs. 27 for major brands
• Different brands reflected different driver
  weights
• Significant factors
• Promotion, Price and price differential,
  Seasonality, Weather, Distribution
• Effort relative to exponential smoothing

8
Model Results for Total Lager Sales
9
Interrelationship Formed

• SCB Lancaster University
• Methodologies analysed
• Wlodek Tych Transfer Function Models
• ACNielsen Promotional Evaluator
• SPSS implementation using Lagged Effects
• Procast
• SCB recognition of benefits of new techniques
• Permanent resource employed

10
Price Focus

• Price - the single most important driver of sales
  volume
• Major cause of forecast error and stock
  shortages/surpluses
• Requirement of tactical and strategic price
  planning
• Series of requirements - advice forecasting
• Comparing price to share (removing seasonality
  aspects)
• By total grocery market and individual customers,
  where EPOS data available
• SKU Brand versus product sector
• SKU Brand versus competitor brand
• Cannibalisation effects

11
Price Focus

• How elastic is the Beer Market
• What is the impact on competitors
• Steal
• Cannibalisation
• Volume
• Price vs. Volume

12
Price Focus

• Identify most profitable Price Level
• Price (RPB) x Volume Profit

Example Brand X in Account when Brand Y _at_ 15.99
X
The Golden Egg
Maximising Profit Contribution
13
Price Elasticity Models

• Use output from exponential smoothing model as
  base
• Recognise confidence interval and implications
• Document assumptions made
• Used for temporary price reductions
• Caution in use as guide for strategic price
  movement
• Need to maintain models reflecting changes in
  market dynamics
• Used with supervision from forecasting team
  currently

14
Cross Elasticity
15
Regression Application

• Price not only factor, need to understand all
  factors that drive beer sales
• dynamic/changing market
• increase in importance of 24Pk
• seasonality/Xmas effect
• Factors considered
• price, competitor pricing, media spend,
  promotion, multibuy, display, feature,
  temperature, seasonality lagged effects, FABs and
  wine effects

16
Methodology

• Link with J.Canduela (PhD Napier University)
• Multiple Regression Techniques
• Three Autoregressive algorithms using SPSS
• Cochrane-Orcutt
• Exact maximum-likelihood
• Prais-Winsten
• Autobox
• Trying to optimise Forecasts whilst keeping
  things easy for the user

17
Current Future

• Methodology running in Multiple Grocer accounts
• Price Promotions
• Strategic Planning
• Infiltrate other segments Wholesale,
  Convenience etc.
• Understand Test different mechanics to evaluate
  optimum performance
• Continue to optimise profitability

18
What Affects Sales ?

• Sales
• Own Promotions Own Trade Activity
• Competitor Promotions Competitor Trade
  Activity
• Own Regular Price
• Own Regular Price vs Competitors Regular Price
• Own TV Advertising
• Competitor TV Advertising
• Distribution Store Effects
• Seasonality
• Random Term

19
Econometric Modelling

• Identifying the relationship between volume sales
  and marketing activity from store-level data

Modeling enables us to understand the impact on
sales of price, promotions and advertising.
20
Being More Sociable

• Unfortunately no samples
• Why are we here I want to learn from others
  why wait?
• Benchmarking my experience
• Compare performance
• Discussion leads to new ideas, new approaches,
  new solutions
• Reduce the number of pitfalls on the way to
  success
• Networking more informal
• Would like to identify other interested parties
  in supply chain
• Agree goals
• Actively involve others
• Meet on regular basis may be electronically