This thesis deals with conversion rate optimization and evaluation of its results. Thesis introduces and compares conversion rate optimization options -- web analytics, heatmaps, user testing and A/B testing. The practical part consists conversion rate optimization of Faculty of Management website and quantification its results.

For an e-commerce business to grow, there are many ways one could try to improve the business in order to gain greater reach and increase sales. One of the main goals of such businesses is to convert as many visitors as possible into customers. Even though many e-commerce businesses already have web analytics tools installed, e-merchants find difficulty in identifying where to start optimizing, what data to extract from analysis reports, and how to make use of such data in order to produce a successful design that will increase the conversion rate. The purpose of this thesis is to (without spending resources on marketing-related factors) guide companies to find a low cost and efficient way to increase the conversion rate by creating well-thought-through designs based on analytic data, qualitative research, and human-computer interaction principles. Google Analytics, a web analytics tool, was used in identifying high-valued pages to optimize and to identify demographics/target groups, while qualitative e-commerce related research was used to shape design-proposal hypotheses. This, along with two A/B tests conducted using Optimizely, is the basis for the guidelines and conclusions. The results of both A/B tests showed an increase in conversions with designs highlighting: evidence of a secure shopping environment, incentives that will attract visitors to buy, and by removing auxiliary navigation elements at the check-out page. The evaluation of the results and its statistical significance was done using both Optimizely’s statistical engine and null hypothesis testing. The increases in conversions were not statistically significant per Optimizely; however, they were significant using traditional statistics. In conclusion, using metrics such as high exit-rates combined with many page views and high revenue-generating pages will allow e-merchants to identify where to start their optimization process. Furthermore, to know what valuable data needs to be extracted, one should seek the data that needs to be inserted into HCI concepts, such as personas and scenarios. This, along with qualitative research allows designers to create well-thought out design-proposals that will potentially lead to an increased conversion rate. / För att få en e-handelsbutik att växa finns det många arbetsområden man kan försöka förbättra för att nå ut till fler samt öka försäljning. Ett av huvudmålen för dessa butiker är att konvertera så många besökare till kunder som möjligt på sin hemsida. Även om många e-handelsbutiker redan har webbanalytiska redskap till sitt förfogande, har många tjänsteleverantörer svårigheter med att fastställa var på hemsidan det skall optimeras, vilken data som ska hämtas från analysrapporter, och hur man använder sig av dessa data för att skapa en lyckad design som kommer öka konverteringsgraden. Syftet med avhandlingen är att, utan marknadsföringsrelaterade investeringar, vägleda företag till billiga och effektiva sätt att öka konverteringsgraden. Detta ska uppfyllas genom att skapa väl genomtänkta designer grundade på analytisk data, kvalitativ forskning, samt människa-datorinteraktions principer. Webbanalysverktyget Google Analytics användes för att identifiera högt värderade sidor att optimera och demografier/målgrupper medan kvalitativ e-handels-relaterad forskning användes för att forma hypoteser kring designförslagen. Detta, tillsammans med två A/B tester som genomfördes med hjälp av Optimizely, är grunden till riktlinjerna och slutsatserna. Resultaten från båda testerna visade en ökning i konverteringar med designer som framhäver; övertygande eller bevis för en säker handelsmiljö, incitament som kommer locka besökare att handla, och genom att ta bort extra navigeringselement vid kassasidan. Utvärdering av resultaten och dess statistiska signifikans gjordes med Optimizelys statistiska motor såväl som egen nollhypotes prövning. Ökningarna av konverteringar var inte statistiskt signifikanta enligt kalkyl från Optimizely, men lyckades nå signifikans enligt traditionell statistik. Sammanfattningsvis, med hjälp av mätvärden så som höga utgångsfrekvenser i kombination med högt antal sidvisningar samt höga intäktsgenererande sidor, kan tjänsteleverantörer nu identifiera var man kan påbörja optimeringsprocessen. För att veta vilken värdefull data man bör extrahera skall man ta reda på vilken data som behövs för att stoppa in i Människa–datorinteraktion (MDI) koncept, som personas och scenarier. Detta, tillsammans med kvalitativ forskning, tillåter webbdesigners att skapa väl genomtänkta designförslag som förhoppningsvis leder till en ökad konverteringsgrad.

<p dir="ltr">This work will introduce a channel conceptualization and possible coding scheme for Graph-and-Symbol (GS) Communication. While Claude Shannon’s mathematical model for communication employed graphs to describe relationships and confusability among traditional time-sequenced signals, little work as been done to describe non-linear communication <i>with</i> graphs where we transmit and receive physical structures of information. The principal contribution of this work is to introduce a mathematical framework for communication with graphs which have symbols assigned to vertices. This looks like a molecule, and so we may think of these messages as coded forms of molecular communication.</p><p dir="ltr">At this time, many problems in this area will (and may remain) computationally intractable, but as the field of graph theory continues to develop, new tools and techniques may emerge to solve standing problems in this new subfield of communication.</p><p dir="ltr">Graphs present two difficulties: first, they contain ambiguities among their vertices and do not have an <i>a priori</i> canonical ordering, and second, the relationships among graphs lack structural regularities which we see in traditional error control coding lattices. There are no Galois fields to exploit over graph-based codes as we have with cyclic codes, for example. Furthermore, the shear number of graphs of order n grows so rapidly that it is difficult to account for the neighborhoods around codewords and effectively reduce communication errors which may occur. The more asymmetric a graph is, the more orderings on symbols it can support. However, asymmetries complicate the computation of channel transition probabilities, which are the cornerstone of all communication theory.</p><p dir="ltr">In the prologue, the reader will be introduced to a new educational tool for designing traditional binary cyclic codes.</p><p dir="ltr">1 through 10 will detail the development of Graph-and-Symbol (GS) Commu- nication to date followed by two example codes which demonstrate the power of structuring information on graphs.</p><p dir="ltr">Chapter 13 onward will review the preliminary work in another area of research, disjoint from the main body. It is included here for posterity and special interests in applying graphs to solving other problems in signal processing. It begins with an introduction of spacetime raythic graphs. We propose a new chamfering paradigm for connecting neighboring pixels which approximates solutions to the eikonal equation. We show that some raythic graphs possess structures with multiple, differing solutions to eikonal wavefront propagation which are essential to the construction of the Umbral Transform. This umbral transform emulates ray casting effects, such as shadows and diffraction within an image space, from a network-flow algorithm.</p><p dir="ltr">This work may be duplicated in whole or in part for educational purposes only. All other rights of this work are reserved by the author, Timothy Arthur Terlep Jr., of Rose-Hulman Institute of Technology, Terre Haute, IN (effective August 2024), and subject to the rules and regulations of the Graduate School of Purdue University.</p><p dir="ltr">Readers may contact the author with any comments and questions at <b>taterlep@gmail.com</b></p>