Chapter 3 BACKUP

Chapter 3 Overview & PowerPoint

Background
  • Forecasted sales used to be based on shipment to customers only (less than ideal)
  • Data was hard to obtain
  • Technology was improving, allowing for more advancements in forecasting

 


Types of Forecasts

    • Forecasts meet the needs of various parts of an organization. The marketing department needs forecasting in order to determine which products or services should be introduced, discontinued, which markets to enter or exit, and which ones to promote. Sales needs forecasting in order to make plans for the company’s future. Supply chain managers use forecasts to make production, procurement, and logical plans for the organization.
    • There are also internal and external demand drivers in a company’s forecast. Internal demand drivers include incentives for salespeople, promotions for consumers, and discounts to trade. External demand drivers unlike internal drivers are uncontrollable. But they can be managed through best practice techniques with structured methods in place to help those in an event that something may occur. This can include unpredictable events that take place in the environment, such as terror attacks and stock market crashes.


General Forecasting Steps

  1. Determine the use of forecast to predict an outcome.
  2. Select which items to be forecasted to be better prepared to possible forecast inaccuracies.
  3. Determine whether the forecast is short, medium, and/or long term, as well as the appropriate “timing planning buckets”.
    • E.g. 30 days or less: daily buckets; 1-3 months out: weekly buckets; 4+ months: quarterly buckets
  4. Select forecasting model/s and methods.
      • Using qualitative and/or quantitative;
      • Considering external factors
        • E.g. customer forecasts, POS data, CPFR, etc.
      • Determining the weight of each factor
  5. Gather data to create forecast. A forecasting software is used to integrate and analyze information. Maintenance and updates will be performed thereafter.
  6. Generate forecasts that involve the use of statistical methods to generate a baseline forecast. The planner will then audit the results and, if needed, try other statistical methods. Management overrides, promotional plans, sales estimations, and externally supplied information will be factored in afterwards.
  7. Validate and implement results. At this point, forecasts are reviewed by several people in the company to obtain high level of accuracy.

Quantitative vs. Qualitative

The qualitative model is typically used when there is not enough data or the situation is somewhat vague. Forecasting with the qualitative method is useful for estimating for new products, services, and technology. Some qualitative methods include knowledge of products, market surveys, jury of executive opinion, and the Delphi method.

  • Knowledge and Intuition of the Products is a forecast that derives from the experience of a forecaster, sales, or marketing that have gathered estimates over a period of time. However, it is important to remember that there are always biases that may come into play with this method.
  • Market Surveys is the process of going directly to the customers to gather information. Focus groups can also be used to ask the perceptions, opinions, beliefs, and attitude towards certain products, services, concepts, advertisements, ideas, or packaging.
  • Jury of Executive Opinion forecasts are determined by managers within an organization that get together and discuss their opinions on what the firm’s future sales may be.
  • The Delphi method is the result of multiple rounds of questionnaires that are sent out by a panel of experts. The anonymous responses are aggregated and shared with the group after each round and the experts are then allowed to modify their answers for each round. This method seeks to reach the correct answer through consensus and is one of the more strategic ways to forecast.

The quantitative models is typically used when there is an ideal amount of historical data to be used for the situation at hand. It is primarily used for current/existing technology products and involves multiple mathematical techniques like time series data models and associative models.

  • Time series forecasting uses a set of evenly spaces numeric data that are obtained by observations of past values and assumes those factors will continue to influence the past, present, and future data. Moving averages and weighted moving average averages are simple and inexpensive ways to predict the future.
  • Associative models forecast based on the assumption the dependent variable has a cause-and-effect relationship with one or more independent variables. Linear and multiple regressions models are used in this case.


Product Life Cycles and Forecasting

It is useful to understand where a product is in its lifecycle when determining whether to rely more on qualitative or quantitative models of forecasting.

  • Introduction: Since there is little history to go off of in the introduction stage, it is better to rely more on qualitative estimates that are generated both internally and externally.
  • Growth: It is very easy to under or over estimate forecasts in this stage, which can have negative effects on costs and service. Both, qualitative and quantitative methods are used to create a blended forecast.
  • Maturity: Since there isn’t much field information needed, you can usually implement simple models. Accuracy tends to improve in this stage.
  • Decline: Sales have a downward trend and demand locations start to shift because the trend is not uniform. Forecasters generally rely more on qualitative than on quantitative methods to run out existing inventory.

Image result for product lifecycle

 

Example: Blockbuster Business Lifecycle 

  • Founded in 1985
  • Netflix goes Public
  • Blockbuster loses 75% of value because of competition
  • Goes bankrupt in 2010

 

 


Time Series Components

    • Trend: An overall upwards or downward pattern due to population, technology, age, culture etc.
    • Cyclical: Repeating up and down movement that are not of fixed period. They are typically affected business cycle, political and economic factors.
    • Seasonal: A seasonal pattern occurs over a fixed and known period and are caused by seasonal factors.
    • Random: Erratic fluctuations due to random variation or unplanned events; often short and non repeatable.

Image result for components of time series


Time Series Models

    • Naive approach: Last period’s actual demand is used as this period forecast demand. For example if January sales were 100, then February forecasted sales will be 100.
    • Moving average: January-March sales are averaged to create an April forecast.
    • Weighted moving average: Typically used when some trend might be present; an average that gives different weights to data at different positions.
      • Weighted moving average forecast for April= .6*March sales + *3 February Sales + *.1January Sales. Weights based off intuition and experience.

Associative Models (More Sophisticated)

    • Linear regression: y= a + bx. Use independent variable (x) to predict dependent variable (y). Able to predict future sales (y) by plugging in the sales budget we plan on using (x). a= the y axis intercept b= slope of the regression line.
    • Correlation
    • Seasonality: Seasonality index may reflect actual seasonal sales of an item (selling more snow shovels in the winter). To create a seasonality index you must do the following:
      • Calculate an average for all item history
      • Average each period’s historical data
      • Divide each period’s average by the overall average
      • Apply the period index to the existing time series or linear regression forecast
      • For example, snow shovel historical quarterly sales as shown
Year Q1 Q2 Q3 Q4
2010 500 100 25 350
2011 550 125 15 400
2012 625 95 30 325
2013 550 130 20 450
Period average 556 113 23 381
Index 2.07 .42 .09 1.42

 

  • Overall average= 268. Index for Q1: 556/268=2.07 Q2: 113/268=.42 ……etc.
  • If we had a quarterly forecast for next year of 100/quarter, we could apply the seasonality index to that forecast.The resulting quarterly forecasts would be Q1= 2.07*100 or 207 shovels; Q2= .42*100, or 42 shovels; Q3= .09*100, or 9 shovels; Q4= 1.42*100, or 142 shovels.

Forecasting Error Measurement

    • Mean Absolute Deviation (MAD) is way to measure the overall forecast error in units over period of time.

  • Mean Squared Error (MSE) is the average of the squared differences between forecasted and actual values 

  • Mean Absolute Percent Error (MAPE) is the most common way for business to measure and control forecast. Calculates the absolute percentage error.

  • Tracking Signal calculates upper and lower control limits to read for errors

 


Accuracy in Relief Demand Planning 

  • Problem: Uncertainties of abrupt relief demand and collaboration of chaotic conditions
  • Solutions:
    1. Multi-echelon supply chain models for disasters
    2. Dynamic facility location and vehicle routing selection
    3. Rescue system management collaboration


Real World Examples

Walgreens
  • During Flu Season, Walgreen analyzes the history of when and where these outbreaks occur combined with current environmental conditions, can predict purchasing demand in the upcoming days and weeks
  • Walgreens tries to understand their customer’s buying behaviors to channel the right products to the right locations


 

Garamba National Park
  • Combats elephant poaching with GIS technology
  • With GIS, African Parks can analyze and monitor:
    1. The movement of 50 elephants that have telemetric collars that allow rangers to track their location
    2. The seasonal movements of pastoralists in the region
    3. How patterns of wildfire are changing at a regional scale
    4. Where trail systems are expanding

Car Demand
  • Finding locations that are more likely to engage in alternate car and ride-sharing options
  • Predicting business opportunities based on emerging trends in precision demographics
    • Neighboring-level earning power
    • Popular shifts
    • Auto ownership fluctuations
  • An auto executive might gauge how and when to diversify by using GIS to map each markets demographics

 


Technology

Software-as-a-Service (SAAS)

Technology in business is a crucial as ever. Competition is constantly increasing, and firms are continually trying to find ways to have an edge on the competition. Technologies to such as Software-as-a-Service are very popular and helpful to assist companies plan for future sales accurately. Loyalty cards and online marketing campaigns play important parts for demand forecasting.

Loyalty Cards 

  • Loyalty cards are a technology that almost every business organization utilizes and are a simple enough concept. Swipe the loyalty card, get a percentage off your price, easy money for the consumer. What the business gets is a wealth of data for incredibly cheap. Swiping a loyalty card when you make a purchase ties all of the information on the card with the purchase. When a traditional sale is made, the company could only track the inventory going out of the store. When a loyalty card is used, a company can track both the inventory and the user to get accurate details on sale trends.

Internet Marketing Campaigns 

  • Another technology that companies utilize daily are internet marketing campaigns. How companies use digital marketing to more accurately forecast demand is how the company can ‘recapture’ data from online ads. As opposed to tv commercials or ads in a paper, companies usually pay ‘per click’ for online advertising. Now companies can tell exactly how much interest a particular product has generated and use that to prepare for initial sale without any product being sold.

 


Powerpoint Presentation Link

Sources: Myerson (2015) Supply Chain Management

https://walgreens.maps.arcgis.com/apps/MapSeries/index.html?appid=40d0763cd3cc42428b26f85202108469

https://www.forbes.com/sites/esri/2018/05/30/combining-ai-and-location-intelligence-to-predict-market-demand/#2a4e5e0038bf

https://www.esri.com/en-us/arcgis/about-arcgis/overview

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3933051/

https://www.forbes.com/2010/05/18/blockbuster-netflix-coinstar-markets-bankruptcy-coinstar_slide.html

https://www.esri.com/about/newsroom/blog/garamba-combats-poaching/

https://www.esri.com/about/newsroom/publications/wherenext/twilight-of-car-ownership/

“https://www.youtube.com/watch?v=muvYChs_FO4”

https://www.hbs.edu/faculty/Publication%20Files/kris%20Analytics%20for%20an%20Online%20Retailer_6ef5f3e6-48e7-4923-a2d4-607d3a3d943c.pdf

“http://demand-planning.com/2018/05/29/how-starbucks-uses-predictive-analytics-and-your-loyalty-card-data/”

http://Sheets.com

“https://blog.bufferapp.com/facebook-ads-guide”

Leave a Reply

Your email address will not be published. Required fields are marked *