Scientific Marketing: Advanced Demand Chain Optimization

Webinar. Applied Scientific Marketing: Machine Learning Methodologies, Advanced Demand Forecasting, Marketing Mix Optimization & Revenue Management

13/01/2021 – 17:00 Madrid - 11:00 Bogota - 10:00 Ciudad de Mexico

Registration Link:

Documents:

?Last PDF version of this document: https://goo.gl/WmHWrm

1. INTRODUCTION

To improve competitive advantage, reduce costs and increase profits, managers seek to integrate the planning of the activities of their value chain, which can be understood as the sum of the supply chain and the demand chain. Due to its complexity, from the conceptual point of view, the problem is divided into two: at the bottom of the chain, related to the transformation of raw materials into commercial products, the best form of industrial production is determined, concentrating the analysis in minimizing the costs of serving an aggregate market in consumer centers; in the upper part, the sale and delivery of commercial products to individual consumers, the analysis focuses on the study of the final customer in order to optimize the price at which the products are sold and at the same time minimize the costs of meet the demand (marketing and sales costs)

The methodologies and technologies used in each case differ in their details, but they are always supported in the fundamental concepts of Advanced Analytics to achieve the optimization of decisions once the random behavior of the decision environment has been studied and characterized. The study of the market is the result of knowledge developed in the behavior of: i) the customers and ii) the competition.

OPCHAIN-DCO (Demand Chain Optimization) is the name given to the set of models developed by DW in order to study the behavior of end customers and the competitors order to understand the process of attention of the demand that is the factor that drives the activities of the supply chain of a company, since it is based on the requirements of the final customers that products that they finally acquire are developed. The projection of demand for different planning horizons is a complex task, since it must consider factors such as: market volatility, competition, the seasonality of the process, unforeseen events, price variations, economic events and the effect of marketing activities, among others. The projection of the demand does not contribute value in itself; the economic value added by the study is capitalized through the decisions made in the marketing and sales process using optimization models.

In OPCHAIN-DCO the focus is on characterizing the aggregate demand of products, by means of mathematical models that quantitatively identify the effect of multiple factors that determine the aggregate behavior of the market. For a better understanding, the market can be segmented into regions, categories, producers and/or brands, among others. The result is the estimation of the aggregate demand, and of all its probabilistic characteristics, based on quantitative relationships between the demand (in quantity and/or in economic value) and the market variables that explain the possible future behavior of the demand in each segment.

OPCHAIN-DCO converts historical records of user sales data into information, its core is the OPCHAIN-DEM model (DEM), whose purpose is to help decision makers probabilistically characterize the dynamic behavior of demand (clients plus competition). This characterization takes the form of dynamic probability distribution functions, conditioned on the marketing and sales actions of the company, and of competition, which will subsequently be used in models that optimize the final decisions leading to the greatest socio-economic benefit in the process of meeting market demand.

It is possible to integrate models that convert the information coming from the analysis of historical demand into economic utility (revenue). The objective of these models is to determine the decisions that maximizes the utility derived from the management of: i) marketing and sales events (marketing-mix), ii) market share policies, iii) pricing policies, iv) inventories and v) sales forces, as a complementary part to the management of the supply chain.

The following table shows the models that make part of OPCHAIN-DCO:

The following diagram presents a summary of the connectivity of the previous models aimed at "managing" the demand in the most convenient way.

2. TYPICAL APPLICATIONS

2.1. GENERAL CONCEPTUALIZATION

The analysis of consumption data based on OPCHAIN-DEM involves building an information system that integrates data from the company's historical consumption records, and from other companies (such as syndicated databases like NIELSEN?) with other companies. sources of information (climate, GDP, ...). From the analysis, customers must be segmented to proceed to estimate mathematical models that explain the demand for each product in each segment.

The results of OPCHAIN-DCO serve as input to other mathematical models that support decisions in multiple functional areas of the organization whether they are demand management or supply management. The central idea is to manage the demand (" demand driven ") to capture most of the social surplus.

2.2. CONSUMPTION PATTERNS CLASSIFICATION

Given the different behavior of the products, the first step in the study of demand is related to the classification of products from the point of view of commercial behavior with a view to determining the "best" statistical model that can be a combination of multiple statistical models. 

There are multiple options for classifying a product: one option is to classify it based on categories imposed by the type of product or type of consumer, for example: size, function, form, conceptual associations, price, utility, cost, volume, among others; by natural (socio-demographic) classifications of the market; Finally, products and consumers can be classified based on the consumption patterns observed in the historical series of demand.

2.3. DEMAND MODELING

OPCHAIN-DEM focuses on obtaining the "best" explanation of the process based on a generic statistical model that includes the effects of the different sources of variation that affect the consumption process.

To carry out this projection, the consumer market model considers the impacts of:

• Elasticity demand – price
• "Pantry loading"
• Lost demand
• Temporary trend
• Periodic seasons
• Seasonality events (commercial initiatives)
• Effects of global exogenous variables
• Effects of regional exogenous variables
• Auto-regressive effects
• Mobile average effects
• Effects of product attributes
• Effects of POS attributes
• Effects of zone attributes.

OPCHAIN-DEM answers quantitatively questions such as:

• What is the immediate effect of the "marketing mix" decisions (2x1 promotions, sharp events, investment in advertising, ...) ?
• What is the impact of periodic seasonings on demand (end of month, fortnight, easter week, day of the week, ...) ?
• What is the impact of the promotions of the subsequent periods ("pantry loading")?
• What is the effect of the characteristics of the product (colors, ...) and the POS in the demand (local size, ...) ?
• What is the natural growth level of the demand, without considering seasonality or marketing events ?
• What is the impact of external variables, such as climate, National or Regional GDP, or population on demand ?
• What is the price-demand elasticity of each reference? Does this elasticity have a seasonal behavior ?

2.4. AGGREGATE DEMAND FORECAST

 The aggregate demand projection of a sector, an area, a manufacturer, a category, and / or a channel, or combinations of the above, has two alternatives: Bottom-Up and Top-Down.

In the Bottom-Up approach, the projection of aggregate demand is made at the lowest level of the grouping (atom level), which implies having the capacity (data and calculation power) to estimate a statistical model for each series of demand (product-POS). In the Top-Down approach the projection is made at the level of aggregate demand and then it is disaggregated at the atom level, for which it is required to have disaggregation rules, which is typically not evident.

Conventionally it is expected that the aggregate demand has less volatility than the demand of the atoms and for that reason, its projection must be more "accurate" than the aggregation of the individual demands, this is fulfilled when the correlation of it is neutral or negative.

 The Bottom-Up focus allows to detect individual effects of atoms that are lost with the Top-Down approach. OPCHAIN-DEM can work under either of the two approaches.

2.4. MATHEMATICAL METHODOLOGIES

The methodologies available to deal with the characterization of demand are: i) Advanced Probabilistic Models (APM), ii) Machine Learning (ML), and iii) Artificial Neural Nets (ANN). Any of these methods (or mixes of them) may be the "best"; however, should be considered the integration between demand and decision-making optimization models based on Mathematical Programming (MP). In these cases, APM and ML are the simplest way to integrate since they are based on algebraic formulas that can be incorporated in MP optimization models. For the case of APM, the algebraic formulas can be incorporated into the models MP; in the case of ML, the Support Vector Machines (SVM) can be integrated into the MP.

The OPCHAIN-DEM approach is oriented to characterize the demand using models APM and/or ML that can be formulated as MP optimization models. OPCHAIN-DEM model is based on an equations simultaneous model (linear or non-linear) that is solved with an optimization technology; the objective function may be: i) minimization of the squares of errors (APM), ii) maximization of likelihood function and iii) and minimization of classification errors (ML).

This approach facilitates to include restrictions on the parameters to bring coherence to the estimation process.

3. ADVANCED ANALYTICS SOLUTIONS

3.1. DECISION-MAKING & CONTROL VARIABLES

The real goal of characterization of demand models is to support decision-making of the responsible for marketing and sales. To make this possible, is required to include in the model exogenous variables that allow the decision-maker to try to influence demand (demand driven) to maximize its revenue.

It is necessary that the demand model can measure the impact on demand of the exogenous variables of control in such a way to include its effect on the optimization models. This means that the forecast models based on time series, as S-ARIMA, may not provide information that requires the decision-maker.

3.2. MARKET SHARE OPTIMIZATION

The characterization of the market share (MS) aims to determine which are the variables that explain the behavior of the MS of a manufacturer (dependent variable), or of a brand, as a function of the characteristics of: market trends; the manufacturer's decisions; the decisions of the competition; and consumer preferences.

Based on the estimation of the elasticity of MS, OPCHAIN-MS can answer:

• What is the impact on the MS of the company's marketing initiatives?
• What is the impact on the MS of competitive marketing initiatives? 

This study can be carried out based on data from syndicated databases such as NIELSEN?. They integrate information from the sales of "all" participants in the market.

3.3. MARKETING MIX OPTIMIZATION

OPCHAIN-MMO converts the information from the analysis of historical sales records into economic profit (revenue, $). Its objective is to determine the decisions that maximize the utility derived from the management of marketing and sales activities; respecting market restrictions, assigned budgets and business rules.

For the optimization of the marketing mix, in addition to the measurement of the impacts of the commercial initiatives (2x1, 10% discount, ...), the impacts that the resources assigned to the clients in sales can be measured, such as: prospectors, promoters, vendors, pre-sellers, active tele-sales, passive tele-sales, internet, shelves, refrigerators, financing models, customer training courses, … .

 MMO answers questions such as:

• What events should be carried out in each period, at each point of sale ?
• What is the cost of marketing initiatives ?
• What is the real increase in demand derived from these initiatives ?
• How much are the additional income due to marketing activities ?
• What should the advertising budget be? And how should it be distributed ?
• How should the sales resources be distributed ?

3.4. REVENUE MANAGEMENT

OPCHAIN-POP (Pricing OPtimization), also known as Revenue Management, is integrated by a set of mathematical programming models by means of which problems oriented to establish pricing policies are solved, that is:

• Price lists
• Prices throughout the life of the product
• Prices to use in an auction event

The objective of mathematical modeling is to determine the best mix of product prices, so as to maximize utility, considering factors such as:

• Seasonality of demand per product
• Elasticity of demand by segments
• Costs per product unit
• Discounts offered by producers
• Fixed order costs
• Variable order costs
• Location of inventories

POP answers questions such as:

• What is the price policy, in each period of the product life cycle, at each point of sale, that maximizes total income ?

The traditional sector that have applied RM, for many decades, are: transport (airline, rail, trains, cruise, cargo…), hospitality (hotel, , tour operators, clubs, restaurants, …), car rental; the non-traditional sectors, that now are applying RM are: energy, broadcast, healthcare, manufacturing, apparel, …

RM models vary according to each type of business, and therefore requires a specific model to each of them.

3.5. SUGGESTED ORDER OPTIMIZATION

The function of OPCHAIN-OPS is to assign the portfolio of products that each "seller" offers to each client (store) in the short term (daily), based on the historical consumption of the client and the availability of inventories. Examples of vendors: a human seller, a car dealer, a seller machine, …

OPS answer questions such as:

• What are the goals that each "sales person" should have ?
• What products should a "seller" offer at each visit ?
• What is the optimal load that a vehicle that makes sales en route ?

4. OPCHAIN-DCO: MODELING THE CUSTOMER

4.1. MARKOVIAN PROCESS

 4.2.   CUSTOMER LIFETIME VALUE (CLV)

 4.3.   CHURN MODELING & OPTIMIZATION

4.4.   RECENCY FREQUENCY AND MONETARY (RFM) MODELING

 4.5. INBOUND CONTACT CENTERS OPTIMIZATION

5. TECHNOLOGY PLATFORM

5.1. CONCEPTUALIZATION

 OPCHAIN-DCO provides the solution to a wide range of strategic, tactical and operational problems considering many aspects of the different types of problems and its components. OPCHAIN-DCO can be used to support consultancy work, end customer studies, or in the "cloud" on an optimization server to directly support decision-making processes.

OPCHAIN-DCO was implemented using OPTEX Optimization Expert System, it can produce program algorithms in various optimization technologies (GAMS, AMPL, AIMMS, C-GUROBI, … ). This implies that it inherits all the characteristics of OPTEX. For more information: https://goo.gl/S2TV8T.

5.2. ALGORITHMS

 OPTEX separates completely the mathematical models of optimization technologies, storing the models in normalized database, allows the user to interact with OPCHAIN-DCO models to update the equations according to the changes in its modeling environment. OPCHAIN-DCO may include services to design, to implement, to support and to solve quickly optimization models in multiples optimization technologies like GAMS, IBM ILOG CPLEX OPTIMIZATION STUDIO, MOSEL, C, AMPL, GMPL … .

Next image presents a GAMS and a C computer programs generated by OPTEX.

5.3. INFORMATION SYSTEMS

 As a complement to other computer systems that require organizations, OPCHAIN-DCO architecture is open (the user knows the data-model) to facilitate their integration with other solutions and/or existing technological tools in organizations (ERP, TMS, WMS, GIS).

5.4. CLIENT-SERVER ARCHITECTURE

OPCHAIN-DCO can operate in a single computer or in a Wide Area Network (WAN) using reference standards intranet/internet. complies with the coordination of tasks and communication between operational modules of OPCHAIN-DCO on different computers on the network. The optimization server facilitates the use of the licenses of the optimization technologies.

6. ABOUT DECISIONWARE

DECISIONWARE (DW) maintains permanent lines of research methodologies large scale optimization and modeling decision problems. Its "know how", through its professionals, has more than forty years working on solving real problems using mathematical programming methodologies.

 OPCHAIN-DCO is a suite of mathematical optimization models, designed and produced by DecisionWare includes solutions to fully optimize multi-business supply chains. Its components have been implemented in many countries and customers in its open structure allow communication with ERP transactional systems. For More information on [email protected]

The following topics may be complementing the information about DW:

1.     DW has developed the OPCHAIN-SCO oriented to optimization of industrial supply chain

https://www.dhirubhai.net/pulse/large-scale-optimization-applied-supply-chain-smart-theory-velasquez/

 2.   The course ADVANCED ANALYTICS & OPTIMIZATION APPLIED TO SUPPLY AND DEMAND CHAINS are related with all topics including in this document (https://goo.gl/EAXTCg)

 3.   The course OPTEX Optimization Expert System include lessons related with large-scale optimization technologies and included a license of OPTEX: https://goo.gl/sAai9T

 4.   More information about:

• OPTEX Optimization Expert System: https://goo.gl/S2TV8T.
• GDDP (Generalized Stochastic Dual Dynamic Programming): https://goo.gl/xVjHCn 

 5.   DW has developed the OPCHAIN-E&G oriented to the electricity and natural gas sectors.

https://www.dhirubhai.net/pulse/electricity-natural-gas-advanced-supply-chain-jesus-velasquez/

6. DecisionWare Experience (https://goo.gl/RkwdeY). It contains a brief description of the main projects that DW has done.

7.   The catalogue of OPCHAIN Mathematical Models presents the models made by DW using OPTEX. (https://goo.gl/3EP9j9)

 8.   Customers can access the OPCHAIN-DCO mathematical models in the following forms:

• On premise, in this case the software OPCHAIN-DCO is installed and runs on the servers indicated by the customer. The software can be sold or leased to the customer.
•  On demand, in this case the customer has access to OPCHAIN-DCO software on a server of DW in the cloud (cloud). The software may also be rented by periods (years, months, semesters, weeks). The databases of the customer may be in a customer server.
•  As a service, through a professional services contract in which DW assumes the responsibility to carry out the agreed work.
• Software Factory, the customer delivers to DW the formulation of the mathematical model and test data and DW delivers to the customer a decision-making support system composed of: i) source of the mathematical model in the selected optimization technology (GAMS, IBM CPLEX, MOSSEL, AMPL, AIMMS,...), ii) data model of information system and iii) test run to probe the correct functioning of the model.

 9.   DW is interested in supporting R+D (Research and Development) in universities, facilitating the know-how accumulated in OPCHAIN-DCO for thesis or for research projects.

 11. DW is looking for business partners, strategic partners and/or technological partners to expand its business around the world.