Since the 1990s, facing increasing competition and mass customization, many companies including Dell have chosen to adopt the assemble-to-order (ATO) model in order to increase products offering and reduce the life cycles of products. Inventory management is a key challenge for ATO systems, in particular determination of inventory replenishment levels without full demand information, component allocations based on available component inventories, and realizations of product demands. ATO systems are usually modeled as a two-stage stochastic integer program. However, such programs are typically hard to solve, especially for stochastic integer nonlinear programs used for the joint optimization. In this thesis, we describe two ATO models proposed by Ackay and Xu (2004) and by Huang (2014). Both models include a nonlinear term in the right hand side of the inventory availability constraints. We discuss the techniques used to linearize the original problem and to estimate the impact of the linearization. In addition, we investigate another key element of ATO systems called component commonality used to reduce inventory costs. An extensive literature review regarding component commonality is provided. / Thesis / Master of Science (MSc)

We present and study the current research trends and advances for optimizing assemble-to-order systems with stochastic lead times. Assemble-to-order systems are widely used in several industries, where end products consist of multiple components and are customizable. Manufacturers can offer greater flexibility to their customers by delaying the assembly of the components until after the order is placed. This means they do not keep an inventory of end-products. Only components can be kept in the inventory and assembled upon demand. We compare and contrast different formulations and the features of assemble-to-order systems studied in the literature including system configuration, replenishment policy, lead time type, and demand distribution. Single-period systems, periodic-review systems, and continuous-review systems are considered. Current assumptions, formulations, solutions, and challenges are discussed. / Thesis / Master of Science (MSc)

We study a periodic review assemble-to-order (ATO) system introduced by Akcay and Xu (2004) which jointly optimizes the base stock levels and the component allocation with an independent base stock policy and a first-come- first-served allocation rule. The formulation is a non-smooth and thus theoretically and computationally challenging. In their computational experiments, Akcay and Xu (2004) modified the right hand side of the inventory availability constraints by substituting linear functions for piece-wise linear ones. This modification may have a significant impact on low budget levels. The optimal solutions obtained via the original formulation, i.e., the formulation without modification, include zero base stock levels for some components and thus indicate a bias against component commonality. We study the impact of component commonality on periodic review ATO systems. We show that lowering component commonality may yield a higher type-II service level. The lower degree of component commonality is achieved via separating inventories of the same component for different products. We substantiate this property via computational and theoretical approaches. We show that for low budget levels the use of separate inventories of the same component for different products can achieve a higher reward than with shared inventories. Finally, considering a simple ATO system with one component shared by two products, we characterize the budget ranges such that either separate or shared inventory component (i.e., component commonality) is beneficial. / Thesis / Doctor of Philosophy (PhD)

Lowering the degree of component commonality may yield a higher type-II service level for a periodic review assemble-to-order system that aims to maximize reward. This is achieved via separating inventories of all the shared components for different products. We investigate the optimal bill-of-materials structure for two-product assemble-to-order systems with arbitrary number of components. The inventory of a shared component can be separated or common between different products. We show that an optimal bill-of-materials can be characterized between the following two extremal configurations: either two products share all common components, or they do not share any common component. / Thesis / Doctor of Philosophy (PhD)

Nous analysons un type W de système de l’Assemble-à-commande avec des délais de livraison aléatoires, l'arrivée aléatoire de la demande et des ventes perdues, en temps continu. Nous formulons le problème en tant que processus de décision Markov à l'horizon infini. Nous nous éloignons de l'approche standard en caractérisant une région de l'espace d'état où toutes les propriétés de la fonction de coût tiennent. Nous caractérisons la politique optimale dans cette région. En particulier, nous montrons que, dans l'intérieur de la région récurrente, les composants sont toujours produits. Nous caractérisons également la politique d'allocation de composants optimale qui spécifie si une demande de produit arrivant devrait être remplie. Notre analyse révèle que la politique d'allocation optimale est contre-intuitive. Par exemple, même lorsqu'un produit domine l'autre, en termes de coût/taux de vente perdue, sa demande peut ne pas avoir une priorité absolue par rapport à la demande de l'autre produit. Une telle caractéristique n'a pas été observée dans de nombreux paramètres intégrés de production/inventaire où l'allocation d'inventaire suit une priorité fixe pour satisfaire les exigences. Nous montrons également que la structure de la politique optimale reste la même pour les systèmes à production par lots, les temps de production répartis par Erlang et la demande de produits non unitaire. Enfin, nous proposons des heuristiques efficaces qui peuvent être utilisées comme substitut à la politique optimale ou peuvent être utilisées comme une politique de départ pour les algorithmes communs utilisés pour obtenir une politique optimale dans le but de réduire leur temps de calcul. / We analyze a W-configuration assemble-to-order system with random lead times, random arrival of demand, and lost sales, in continuous time. We formulate the problem as an infinite-horizon Markov decision process. We deviate from the standard approach by first characterizing a region (the recurrent region) of the state space where all properties of the cost function hold. We then characterize the optimal policy within this region. In particular, we show that within the interior of the recurrent region components are always produced. We also characterize the optimal component allocation policy which specifies whether an arriving product demand should be fulfilled. Our analysis reveals that the optimal allocation policy is counter-intuitive. For instance, even when one product dominates the other, in terms of lost sale cost and lost sale cost rate (i.e., demand rate times the lost sale cost), its demand may not have absolute priority over the other product’s demand. Such a feature has not been observed in many integrated production/inventory settings where inventory allocation follows a fixed priority in satisfying demands. We also show that the structure of the optimal policy remains the same for systems with batch production, Erlang distributed production times, and non-unitary product demand. Finally, we propose efficient heuristics that can be either used as a substitute for the optimal policy or can be used as a starting policy for the common algorithms that are used to obtain the optimal policy in an effort to reduce their computational time

Les systèmes assemble-to-order (ATO) peuvent être considérés comme une affectation de ressources multiples qui induit planification de production, satisfaction des contraintes et affectation des stocks. Les systèmes ATO représentent une stratégie de logistique populaire utilisée en gestion de fabrication. En raison de la complexité croissante des systèmes de fabrication d'aujourd'hui, le défi pour les systèmes ATO est de gérer efficacement les stocks de composants et de trouver les décisions optimales de production et d'affectation.Nous étudions un système ATO avec un produit unique qui est assemblé à partir de plusieurs composants. Le système doit répondre à une demande non seulement du produit assemblé, mais aussi des composants individuels. Nous considérons le cas avec seulement des lost sales puis le cas mixte lost sales et backorders avec des temps de production suivant des lois de type exponentiel et une demande sous forme de loi de Poisson. Nous formulons le problème comme un Processus de décision markovien (MDP), et nous considérons deux critères d'optimalité qui sont le coût actualisé et le coût moyen par période. Nous caractérisons la structure de la politique optimale et étudions l'impact des différents paramètres du système sur cette politique. Nous présentons également plusieurs heuristiques pour le cas lost sales et le cas mixte lost sales et backorders. Ces heuristiques fournissent des méthodes simples, mais efficaces pour contrôler la production et l’affectation des stocks du système ATO / Assemble-to-order (ATO) systems can be regarded as a multiple resource allocation that induces production planning, requirements fulfilling and inventory assignment. ATO is a popular strategy used in manufacturing management. Due to the increasing complexity of today’s manufacturing systems, the challenge for ATO systems is to efficiently manage component inventories and make optimal production and allocation decisions. We study an ATO system with a single product which is assembled from multiple components. The system faces demand not only from the assembled product but also from the individual components. We consider the pure lost sales case and the mixed lost sales and backorders case with exponential production times and Poisson demand. We formulate the problem as a Markov decision process (MDP), and consider it under two optimality criteria: discounted cost and average cost per period. We characterize the structure of the optimal policy and investigate the impact of different system parameters on the optimal policy. We also present several static heuristic policies for the pure lost sales and the mixed lost sales and backorders cases. These static heuristics provide simple, yet effective approaches for controlling production and inventory allocation of ATO system

Sammanfattning Kurs: Examensarbete i Logistik för Civilekonomprogrammet, 30 hp, 4FE19E.  Författare: Jakob Bergman och Jim Andersson. Examinator: Peter Berling Handledare: Peter Berling  Medbedömare: Hana Hulthén Titel: Saldoavvikelser i producerande företag, en fallstudie på Getinge Disinfection AB. Bakgrund: En välfungerande lagerhållning är en väsentlig grundsten för ett företags framgångar. I takt med att lagerhållningen blivit mer automatiserad och effektiviserad ställs högre krav på att informationen om lagersaldon stämmer överens med verkligheten. Saldoavvikelser är ett utbrett problem som kan påverka många olika delar av en verksamhet, såsom felaktiga underlag för inköp, den dagliga driften och företagets servicegrad. Fallföretaget Getinge Disinfection AB upplever främst att saldoavvikelser på företaget får effekt på den dagliga driften, något som denna studie ämnar undersöka. Syfte: Studien syftar till att identifiera konsekvenser och orsaker till saldoavvikelser på Getinge Disinfection AB samt ge förbättringsförslag som kan reducera saldoavvikelser på företaget. Metod: Denna studie har en kvalitativ forskningsstrategi. Vidare är studiens forskningsdesign en fallstudie då den endast undersöker ett fall ingående och detaljerat. Resultat: Studien fann att saldoavvikelser kan uppstå i alla steg i den studerade processen, med varierande primära orsaker i de olika stegen i processen. Bland de konsekvenser som funnits är de främsta icke-värdeskapande aktiviteter, opålitliga underlag för inköpsbeslut, OOS och ökade ledtider. Majoriteten av orsakerna grundar sig i den mänskliga faktorn, det vill säga att anställda inte följer rutiner eller gör misstag i processen. Dessa misstag kan ske i olika aktiviteter i processen och det finns olika primärorsaker beroende på var i processen misstaget sker. Vidare har författarna tagit fram förbättringsförslag för att effektivisera Getinge Disinfection ABs produktionsprocess. De mest lämpliga förbättringsförslagen är att skapa förståelse om saldoavvikelser bland de anställda, införa Cycle counting, digitalisera olika aktiviteter och utvärdera det nuvarande artikelsortimentet. Nyckelord: Saldosäkerhet, Saldoavvikelser, Producerande företag, Medicinteknik, Life Science, Assemble to Order, ATO, Cycle Counting, RFID, Streckkoder. / Abstract  Course: Degree project in Logistics, the Business Administration and Economics Programme, 30 credits, 4FE19E.    Authors: Jakob Bergman and Jim Andersson. Examiner: Peter Berling.  Tutor: Peter Berling. Co-judger: Hana Hulthén Title: Inventory record inaccuracies in manufacturing companies, a case study on Getinge Disinfection AB. Background: Effective inventory management is of considerable importance for most successful businesses. Without it, companies would be hard-pressed to meet the demands of consumers that place a high value on fast deliveries and product availability. As a result, businesses have turned to increased automation of their inventory management. This automation relies heavily on information such as inventory records being accurate and up to date. Inventory record inaccuracies result in purchases or decisions being made on incorrect assumptions of the companies’ actual inventory status, which has a negative effect on many aspects of the business. However, most studies on this subject focus on inventory record inaccuracies in a retailing context, whereas this study focuses on a large manufacturing company (Getinge Disinfection AB) using an Assemble to Order model. Purpose: The purpose of this study is to identify the different causes and consequences of inventory record inaccuracies (IRI) at Getinge Disinfection AB. Furthermore, the study aims to offer proposals on how to improve inventory record accuracy at Getinge Disinfection AB.    Methodology: This paper adopts a qualitative research strategy with a case-study approach.  Findings: The study concluded that inventory record inaccuracies can arise in all areas of the studied process, with different primary reasons for different stages of the process. Most of the causes of IRI can be attributed to the human factor, such as incorrect identification of the items, unregistered movements of goods, and transaction errors. The consequences of (IRI) were mainly the resulting non-value adding activities, purchasing decisions being made based on unreliable inventory data, stockouts and longer lead times. The study also gives suggestions on how Getinge could reduce these inventory record inaccuracies. Improvement measures such as Cycle counting, bar coding of the items and locations and increased staff training are discussed. Keywords: Inventory Record Accuracy, Inventory Record Inaccuracy, Manufacturing, Health Care, Life Science, Assemble to Order, ATO, Cycle Counting, RFID, Bar coding.

In an attempt to reduce costs and improve customer satisfaction, manufacturers have been adopting strategies such as Delayed Product Differentiation (DPD) while managing broader product lines. In this study, first a general framework on DPD is formed in the light of basic articles in the literature. The vanilla box assembly process which is a special form of modular design type of DPD is modeled and analyzed. In the vanilla box assembly process, inventory is stored in a special form of semi-finished products, called vanilla boxes, that can serve more than one final product. We model the vanilla box assembly process considering the costs of inventory and unsatisfied demand under the capacity limitations, stochastic demand and bill of material requirements. We formulate the model as an extensive form of stochastic integer program in which stochastic demand is modeled using a set of demand scenarios each of which is assigned a probability of occurrence. The model is solved as a standard integer programming model that minimizes the expected value of the objective function. The impact of product demand scenarios, common component levels, shortage penalty cost to holding cost ratio levels and capacity restrictions on the total cost and fill rates is studied. We compare the performance of vanilla box assembly process to assemble-to-order process and provide insights on their performances. Computational results indicate that the vanilla box assembly process is a promising alternative to the assemble-to-order process in most of the problem instances.

This report is the result of a 20-points project at the University of Jönköping. The project was carried out in form of a case study with the object of analysing the value flow at DIAB AB’s confection department in Laholm. The aim of this project is to submit proposals to the production management on how to increase the efficiency of the production flow at the confection department and reduce the capital accumulation in form of products in work. The information in this report is gathered from interviews, observations and measurements. Furthermore, a literature study was carried out in view to find suitable theories when analysing present as well as future suggested production conditions. This report is based on four main questions: • What does the existing process of the value flow for the most produced product family look like? • How does the process of the value flow for GS perform considering the through-put-time? o How long through-put-time does a representative product of the GS-family have? o How long is the value- and no value adding time for that product along its production flow? • Which production related disturbances and cost prompters exist in the present value flow process? • How could the process of the value flow for GS be made more efficient, less persistent to disturbances and more competitive? The existing process of the value flow for the most producing product family has been mapped and is illustrated in appendix 3. For now, the process includes nine working stations along the production chain. The through-put-time of a representative GS-product is according to my survey 18,5 days. The value adding time is only 16,1 minutes, that is 0,061 % of the entire through-put-time. The remaining time, in other words the no value adding time, is 440 hours and it represents mainly storage and transport of products. The representative production disturbances and cost prompters that characterise the process of the value flow contain material related disturbances, a high number of long shifts, long storage time prior to the customer order point and with that, high capital accumulation and finally unnecessary transports. Improvement proposals aim to increase the efficacy of the process of the value flow and reduce the capital amounts by shifting from the present production strategy involving manufacturing towards order (TMO) to assembling towards order (MMO). In order to make this possible a semi-manufactured storage will be introduced after the standard confection which will represent the new decoupling point. The production at the standard confection will then occur according to the semi-manufactured storage. The standard confection should produce in larger aggregated order quantities based on prognosis in order to benefit from the advantages of economy of scale and the production must proceed in a continuous flow according to the FIFU-system (First In First Out). In addition to that, the special confection must produce according to a pull-system and only when the customer makes a request. The tact-time of the GS products should constitute a limit for all the cycle times along the production chain, both on the standard- and special confection. This is partly due to creating a constant and balanced production flow which enables short through-put-time and partly due to avoiding in-between-storage as a result of various bottlenecks.

Les systèmes assemble-to-order (ATO) peuvent être considérés comme une affectation de ressources multiples qui induit planification de production, satisfaction des contraintes et affectation des stocks. Les systèmes ATO représentent une stratégie de logistique populaire utilisée en gestion de fabrication. En raison de la complexité croissante des systèmes de fabrication d'aujourd'hui, le défi pour les systèmes ATO est de gérer efficacement les stocks de composants et de trouver les décisions optimales de production et d'affectation.Nous étudions un système ATO avec un produit unique qui est assemblé à partir de plusieurs composants. Le système doit répondre à une demande non seulement du produit assemblé, mais aussi des composants individuels. Nous considérons le cas avec seulement des lost sales puis le cas mixte lost sales et backorders avec des temps de production suivant des lois de type exponentiel et une demande sous forme de loi de Poisson. Nous formulons le problème comme un Processus de décision markovien (MDP), et nous considérons deux critères d'optimalité qui sont le coût actualisé et le coût moyen par période. Nous caractérisons la structure de la politique optimale et étudions l'impact des différents paramètres du système sur cette politique. Nous présentons également plusieurs heuristiques pour le cas lost sales et le cas mixte lost sales et backorders. Ces heuristiques fournissent des méthodes simples, mais efficaces pour contrôler la production et l'affectation des stocks du système ATO