Optimizing Shelf Space Allocation In Grocery Retail
Aalto UniversitySchool of ScienceMaster’s Programme in Mathematicsand Operations ResearchElla WarrasOptimizing Shelf Space Allocation inGrocery RetailMaster’s ThesisEspoo, May 22, 2019Supervisor:Advisor:Professor Harri EhtamoTuomas Viitanen D.Sc. (Tech.)The document can be stored and made available to the public on the open internetpages of Aalto University. All other rights are reserved.
Aalto UniversitySchool of ScienceMaster’s Programme in Mathematicsand Operations ResearchABSTRACT OFMASTER’S THESISAuthor:Ella WarrasTitle:Optimizing Shelf Space Allocation in Grocery RetailDate:May 22, 2019Pages: viii 72Major:Systems and Operations ResearchCode: SCI3055Supervisor:Professor Harri EhtamoAdvisor:Tuomas Viitanen D.Sc. (Tech.)Grocery retail is a competitive industry with high sales volumes and low profitmargins, which makes managing costs and optimizing processes especially important. Store and warehouse labor costs constitute a large part of the retail coststructure, and it is also an area where large savings can be obtained by optimizingdifferent processes. Optimizing the use of shelf space can reduce the amount oftime the employees have to spend bringing stock from the backroom storage tothe shelf. Other benefits of an optimized allocation of shelf space include reducedlost sales and overall increases in customer satisfaction.The goal of this thesis is to find a way to divide the available shelf space between agiven set of products so that the need for restocking the shelves is reduced and theopportunity cost in the form of lost sales is minimized. This approach is differentfrom the existing methods in literature, many of which focus largely on the spaceelasticity of the demand. In this thesis, the shelf space allocation problem isformulated as an optimization problem, where the function to be minimized isthe expected quantity of lost sales. The main constraint is the available shelfspace.The optimization problem is solved using the simulated annealing algorithm,and different variations of the algorithm are compared. The algorithm performswell with a linear cooling schedule and a static step size of 1. Good results arealso obtained with a logarithmic cooling schedule, when the control parameter ischosen carefully. Using a method known as thermodynamic simulated annealingdid not result in improvements for the test cases. In all of the variations, theselection of the initial temperature was found to have a significant impact.The simulated annealing algorithm is a valid option for solving the shelf spaceallocation problem. There are variations of the algorithm that are suitable fordifferent situations, and by optimizing the values of the different parameters onecan improve the results. Further research is still needed before using these resultsin real-life applications.Keywords:space planning, shelf space allocation, simulated annealing,integer optimization, nonlinear knapsack problemLanguage:Englishii
Aalto-universitetetHögskolan för teknikvetenskaperMaster’s Programme in Mathematicsand Operations ResearchSAMMANDRAG AVDIPLOMARBETETUtfört av:Arbetets namn:Ella WarrasOptimering av hyllutrymmesallokering inom dagligvaruhandelnDatum:22 maj 2019Sidantal: viii 72Huvudämne:Systems and Operations ResearchKod:SCI3055Övervakare:Professor Harri EhtamoHandledare:Teknologie doktor Tuomas ViitanenDagligvaruhandeln är en bransch med hård konkurrens, höga försäljningsvolymeroch låga vinstmarginaler, vilket betyder att det är särskilt viktigt att hålla kostnaderna under kontroll och optimera processerna. Arbetskraftskostnaderna i butiker och lager utgör en stor del av detaljhandelns kostnadsstruktur, och detfinns även stor potential för besparingar inom det området. Genom att optimeraanvändningen av hyllutrymme är det möjligt att minska på tiden de anställda ärtvungna att använda på att föra varor från lagret till hyllan. Andra fördelar är enminskning av den förlorade försäljningen och en allmän ökning i kundnöjdheten.Målet med detta diplomarbete är att hitta ett optimalt sätt att fördela dettillgängliga hyllutrymmet mellan en given uppsättning produkter så att behovet att fylla på hyllorna minskar och möjlighetskostnaderna i form av förloradförsäljning minimeras. Denna prioritering skiljer sig från befintliga metoder i litteraturen, varav många fokuserar starkt på efterfrågans utrymmeselasticitet. I detta arbete formuleras hyllutrymmesallokeringsproblemet som ett optimeringsproblem, där funktionen som minimeras är den förlorade försäljningens väntevärde.Huvudsakliga bivillkoret är det tillgängliga hyllutrymmet.Optimeringsproblemet löses med hjälp av metoden simulerad glödgning, och olikavarianter av algoritmen jämförs. Algoritmen presterar väl med en linjär nedkylningsfunktion och en konstant stegstorlek på 1. Goda resultat nås även med enlogaritmisk nedkylningsfunktion, då kontrollparametern väljs noggrant. En metod som kallas termodynamisk simulerad glödgning ledde inte till förbättringar iresultaten för testfallen i denna studie. I alla varianter av algoritmen hade valetav starttemperatur en betydande inverkan.Simulerad glödgning är ett fungerande alternativ för att lösa hyllutrymmesallokeringsproblemet. Det finns varianter av algoritmen som lämpar sig för olika situationer, och genom att optimera värdena på de olika parametrarna kan resultatenförbättras. Fortsatt forskning behövs ännu innan dessa resultat kan användas förverkliga tillämpningar.Nyckelord:hyllplanering, allokering av hyllutrymme, simuleradglödgning, heltalsoptimering, ickelinjärt kappsäcksproblemSpråk:engelskaiii
Aalto-yliopistoPerustieteiden korkeakouluMaster’s Programme in Mathematicsand Operations ResearchDIPLOMITYÖNTIIVISTELMÄTekijä:Työn nimi:Päiväys:Pääaine:Ella WarrasPäivittäistavarakaupan hyllytilan allokoinnin optimointi22. toukokuuta 2019Sivumäärä: viii 72Systems and Operations Re- Koodi:SCI3055searchValvoja:Professori Harri EhtamoOhjaaja:Tekniikan tohtori Tuomas ViitanenPäivittäistavarakaupan alalla kilpailu on kovaa ja myyntimäärät ovat suuria, mutta voittomarginaalit pieniä, minkä vuoksi kulujen hallinta ja prosessien optimointi on erityisen tärkeää. Myymälöiden ja varastojen työvoimakulut muodostavatsuuren osan vähittäiskaupan kustannusrakenteesta, ja se on myös osa-alue, jollavoidaan saavuttaa suuria säästöjä optimoimalla eri prosesseja. Hyllytilan käytönoptimointi voi vähentää työntekijöiltä aikaa, joka kuluu tavaran siirtämisessä takahuoneesta hyllyyn. Muita etuja optimoidussa hyllytilan allokoinnissa ovat menetetyn myynnin väheneminen sekä yleinen asiakastyytyväisyyden kasvu.Tämän diplomityön tavoite on löytää optimaalinen käytössä olevan hyllytilanjako annettujen tuotteiden välillä siten, että hyllytystyön tarve vähenee ja vaihtoehtoiskustannukset menetetystä myynnistä laskevat. Tämä lähestymistapa onerilainen verrattuna kirjallisuudesta löytyviin menetelmiin, sillä monet niistä keskittyvät pääosin kysynnän tilajoustoon. Tässä diplomityössä hyllytilan allokointiongelma muotoillaan optimointiongelmana, jossa minimoitava funktio on menetetyn myynnin odotusarvo. Ongelman tärkein rajoite on käytössä oleva hyllytila.Optimointiongelma ratkaistaan käyttämällä simuloitu jäähdytys -menetelmää, jaalgoritmin eri variaatioita vertaillaan. Algoritmi tuottaa hyviä tuloksia lineaarisella jäähdytysfunktiolla ja staattisella askelkoolla 1. Tulokset ovat myös lupaaviakun käytetään logaritmista jäähdytysfunktiota, mutta se vaatii säätöparametrinhuolellista valintaa. Termodynaaminen simuloitu jäähdytys -niminen menetelmäei tuottanut parannuksia testien tuloksiin. Kaikissa variaatioissa alkulämpötilanvalinnalla osoittautui olevan suuri merkitys.Simuloitu jäähdytys on toimiva algoritmi hyllytilan allokointiongelman ratkaisuun. Algoritmista on variaatioita, jotka soveltuvat erilaisiin tilanteisiin, ja tuloksia voi parantaa optimoimalla eri parametrien arvoja. Aiheesta tarvitaan vieläjatkotutkimusta ennen kuin tuloksia voi käyttää tosielämän sovelluksissa.Asiasanat:hyllysuunnittelu, hyllytilan allokointi, simuloitu jäähdytys,kokonaislukuoptimointi, epälineaarinen selkärepun täyttöongelmaKieli:englantiiv
AcknowledgementsThis thesis has been a challenging project, and there are many people thathave helped me along the way. First, I would like to thank RELEX Solutionsfor giving me the opportunity to write this thesis. A special thanks goes tomy advisor Tuomas Viitanen, whose support, dedication and encouragingattitude have been invaluable for the thesis process. I also wish to thank mysupervisor Harri Ehtamo for providing his feedback and expertise.I have had the opportunity to experience and learn so much during thepast years, and I have made many wonderful friends along the way. Thecommunity in Otaniemi and especially Teknologföreningen, my second homeduring these years, have played a large part in my life and I cannot thank allthe people involved enough.A huge thanks also goes to my colleagues at RELEX for supporting me andmaking the workdays fun during the last couple of years. All the discussionsand support have helped me in the process of completing this thesis.Most importantly, I want to thank my friends and family for all their loveand support. Bra faijor, without you the past years would have been muchmore boring, and I hope we have many more adventures together in thefuture! Finally, I want to thank my mom and dad for always supporting mein everything I do.Espoo, May 22, 2019Ella Warrasv
Glossarydays of supplynumber of days the current stock will lastwhen taking into account future demandfacingone unit of a product that is visible on thefront on the shelf or other fixturefixtureany type of shelf or other structure that canbe used for presenting productsheuristicssimple rules applied empirically to find a”good enough” solution quicklylead timetime between order and deliverymacro space planningfloor space planning, decisions about where onthe store map product categories are placedmetaheuristicsmethods that are more general and problemindependent than heuristics, provide a morethorough approachmicro space planningshelf space planning, decisions about where individual products are placed on the shelfplanogramshelf plan in picture form, shows where eachproduct is to be placed on the shelfstock-outwhen the product is sold outvi
ContentsGlossaryv1 Introduction11.1Problem Statement . . . . . . . . . . . . . . . . . . . . . . . .21.2Scope of the Thesis . . . . . . . . . . . . . . . . . . . . . . . .31.3Structure of the Thesis . . . . . . . . . . . . . . . . . . . . . .32 Shelf Space Allocation52.1Assortment Planning . . . . . . . . . . . . . . . . . . . . . . .52.2Floor Space Planning . . . . . . . . . . . . . . . . . . . . . . .82.3Shelf Space Planning . . . . . . . . . . . . . . . . . . . . . . . 102.3.1Replenishment Costs . . . . . . . . . . . . . . . . . . . 122.3.2Space Elasticity . . . . . . . . . . . . . . . . . . . . . . 132.3.3Cross-Space Elasticity . . . . . . . . . . . . . . . . . . 132.4Shelf Space Planning in Literature . . . . . . . . . . . . . . . . 142.5Shelf Space Allocation as an Optimization Problem . . . . . . 172.5.1Objective Function . . . . . . . . . . . . . . . . . . . . 192.5.2Constraints . . . . . . . . . . . . . . . . . . . . . . . . 22vii
3 Solution Algorithm253.1Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2Possible Solution Methods . . . . . . . . . . . . . . . . . . . . 273.3Simulated Annealing . . . . . . . . . . . . . . . . . . . . . . . 293.3.1Algorithm Description . . . . . . . . . . . . . . . . . . 293.3.2Test Setup . . . . . . . . . . . . . . . . . . . . . . . . . 323.3.3Thermodynamic Simulated Annealing . . . . . . . . . . 334 Results374.1Linear Cooling Schedule . . . . . . . . . . . . . . . . . . . . . 384.2Dynamic Neighbor Function . . . . . . . . . . . . . . . . . . . 404.3Logarithmic Cooling Schedule . . . . . . . . . . . . . . . . . . 434.4TSA Cooling Schedule . . . . . . . . . . . . . . . . . . . . . . 434.5Adapted TSA Cooling Schedule . . . . . . . . . . . . . . . . . 474.6Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485 Conclusions505.1Key Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . 505.2Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525.3Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . 54A Complete Test Results60viii
Chapter 1IntroductionIn the world of retail, the combined sales for the global top 250 companiesreached US 4.4 trillion in 2016 (Deloitte, 2018). In the USA, grocery retailers sold US 648 billion worth of products in 2016, and even in Finlandthe grocery market reached EUR18.2 billion in value in 2018 (United StatesDepartment of Agriculture, 2018; Nielsen, 2019). Grocery retail is an enormous industry, but profit margins are relatively small. It was ranked one ofthe least profitable industries in 2017, with a net profit of only 2.2% (Biery,2017). Grocery is also an increasingly competitive industry, and all of thismeans managing costs and optimizing processes is especially important, aseven small improvements can result in huge savings in expenses for the retailer. Research by e.g. Angerer (2006) shows that there is a lot of potentialfor improvement in the fast moving consumer goods industry by using different technological solutions for optimizing the store replenishment process.Store and warehouse labor costs constitute a large part of the retail coststructure, and it is also an area where large savings can be obtained by optimizing different processes. Effective space planning can save time for thestore employees in the shelf stacking process. The savings can come frommany different parts of the process, but one aspect is the shelf space allocation, which, if done optimally, can reduce the amount of time the employeeshave to spend bringing stock from the backroom storage to the shelf. Optimizing the use of shelf space brings many other benefits too, such as reducedlost sales when customers are not met with an empty shelf where their preferred product should be, and overall increases in customer satisfaction whenthe full assortment of products is presented in a clear way, without out-ofstocks.1
CHAPTER 1. INTRODUCTION2A review by Hübner and Kuhn (2012) shows that in the area of retail category management, there is a large amount of high-quality research on howto best manage the space and assortment aspects in the stores. However,that research knowledge has not reached the software solutions that existtoday. Most systems still use simple rules and settings for making space andassortment planning decisions, while the methods found in the literature aremore advanced. There is potential for closer cooperation between the two;practical software solutions can become more intelligent by incorporating research findings, and research studies can benefit from some real-life insightsabout the use cases.1.1Problem StatementIn this thesis, the goal is to formulate a practical method to be used in shelfspace allocation planning. More specifically, the idea is to find a way todivide the available shelf space between a given set of products so that theneed for restocking the shelves is reduced and the opportunity cost in theform of lost sales is minimized.The objective of this study can be formulated as follows:What factors should be taken into account when allocating shelf space between products in a retail store, and how can the optimal allocation be solvedefficiently?In order to answer this question, there are some steps that need to be taken.Firstly, it is important to study the current available methods that have beenused for solving the shelf space allocation problem. After that, the shelf spaceallocation is formulated as an optimization problem, which means decidingwhat a good objective function is, as well as defining the optimization constraints. Then, based on the characteristics of the optimization problem, asuitable solution method needs to be found and implemented. The selectedalgorithm needs to be suited for the different requirements of the optimization problem. Different variations of the algorithm are tested and comparedin order to find the most efficient method for the optimization.
CHAPTER 1. INTRODUCTION1.23Scope of the ThesisThis study is focused on shelf space allocation in retail stores, and specificallygrocery retail. The findings can be applicable to some non-grocery retailstores, but the examples and test data are from the grocery industry. Shelfspace optimization can only be done if the shelf space is, in fact, limited.This is the case in most grocery stores, but excludes some specialty itemssuch as premium class watches or clothing. The space planning decisions ofthose retailers typically do not concern a limited amount of space that needsto be filled, so the approach is quite different.The study is limited to products with previous sales history, so completelynew products that are being introduced are not included in the scope. It is,however, possible to use the sales history of another product as a reference, ifthere is a comparable reference product to be assigned. The products in thisstudy are also assumed to have regular and reasonably frequent deliveries tothe store, since the allocated shelf space is meant to satisfy enough demandso that few refills of the shelf are needed before the next delivery. It couldalso work in some situations with infrequent deliveries, but the main focusof this thesis is the frequently delivered products, such as grocery products.The changes in demand as a function of changes in the amount of space werenot included in the study, although the topic is discussed in Section 2.3 ofthis thesis. Space elasticity was determined to be such a wide topic, that itwas excluded from the scope of the thesis. The same applied to substitutioneffects and cross-space elasticity.1.3Structure of the ThesisThe thesis starts with a chapter on the general background of space andassortment planning, with the key concepts presented. Shelf space planningis specifically discussed in more detail, including some key concepts relatedto replenishment and demand elasticity. After this, a review of relevantliterature on shelf space planning is presented. The chapter is concludedwith a section on the formulation of the shelf space allocation problem as anoptimization problem, with an objective function and constraints.Next, in Chapter 3, a general introduction into optimization is given, afterwhich the different possible solution methods for this type of an optimization problem are discussed. The selected method, the simulated annealing
CHAPTER 1. INTRODUCTION4algorithm, and its advantages are presented in detail, along with descriptionsof the variations that are tested in this study. The results are presented inChapter 4, along with a description of the test data and other details related to the test setup. The chapter is divided into sections for the differentmethods that were tested. In Chapter 5 the key findings of the study arepresented, and some analysis is provided on the quality and practical implications of the results. In the final section, some suggestions are presented forfuture research possibilities.
Chapter 2Shelf Space AllocationIn this chapter, a general overview of space and assortment planning is presented, along with definitions for some key concepts in this area. After that,some previous approaches to solving the shelf space allocation problem arereviewed.In retail stores there are many space and assortment planning aspects toconsider. Assortment planning refers to determining the set of products thatthe store should sell in each category, while space planning covers a widerange of major and mi
chosen carefully. Using a method known as thermodynamic simulated annealing did not result in improvements for the test cases. In all of the variations, the selection of the initial temperature was found to have a signi cant impact. The simulated annealing algorithm is a valid option for solving the shelf space allocation problem.
Quik-Lok Shelving POSITION UPRIGHT Temporarily place one shelf support at any convenient location to properly space uprights. 1 BUILD FROM THE BOTTOM UP Place shelf supports or quick- base (front and back) for lowest shelf only. 2 FIT SHELF OVER SUPPORTS Then add two more supports at the next shelf location, place shelf, and continue up the unit. 3
Crème de Menthe or Get 27 Top Shelf Creme de Menthe Hot Cinnamon Top Shelf Hot Cinnamon Schnapps Parfait Amour Top Shelf Parfait Amour Pimm’s No.1 Top Shelf Summer Cup No.1 Port Wine Top Shelf Ruby Port Southern Comfort Icon Southern Smooth Top Up Liqueur Kit Vermouth Top Shelf Dry Vermouth IF YOU LIKE YOU WILL ENJOY COMMERCIAL EQUIVALENTS
Connect the communication cable between Distribution Shelf and adjacent Rectifier Shelf (cable provided). . Install other controller communications cables (alarm, LAN, etc.) per Rectifier Shelf Quick Start Guide. Controller in Distribution Shelf. ABB Device Comm cable a. Route cable out of the shelf through the Control Cable Bushing. b.
430 allocation to elianto cfd o&m 20,577.32 440 allocation to trillium west cfd o&m 27,267.00 450 allocation to west park cfd o&m 70,008.22 460 allocation to festival ranch cfd o&m 177,790.54 480 allocation to tartesso west cfd o&m 27,809.17 481 allocation to anthem sun valley cfd o&
Space Officer will make recommendations for any major changes in space allocation and remodeling through the Dean/Director/Vice Provost to the Office of Space Management. The Space Officer or committee may request OSM's assistance in analysis of space utilization and allocation at any stage of the process. (See Appendix
Adoramus Te Clement/Leavitt In Library Shelf 1 9 SATB Adoramus Te, Christe Palestrina In Library Shelf 1 24 SATB Africa Lawson In Library Binder 1 SATTBB Africa Call To Peace Liebergen In Library Shelf 1 16 SATB Aftonen Alfven In Library Shelf 1 29 SATB Agnus Dei Hassler In Library Shelf 1 27 SATB
Manual ice tray with lid (non-Ice & Water models only) . Large door shelf Mid door shelf Large door shelf Dairy shelf Mid door shelf freezer compartment Ice maker (top left of freezer) (Ice & Water models only) Ice storage bin with sco
Use the English phonemic alphabet page, which you find at the beginning of good dictionaries, as a guide to pronouncing new words. Effective English Learning ELTC self-study materials Tony Lynch and Kenneth Anderson, English Language Teaching Centre, University of Edinburgh 2012 9 3. Don't forget to learn the word stress of a new word. Every English word has its own normal stress pattern. For .
