Kasutaja tajutava laadimise mõõdikute mõju konversioonimäärale mõõtmine ettevõtte TransferWise kontekstis
Abstract
Kasutaja kogemus veebilehe laadimisel võib mõjutada ärimõõdikuid. Üks nendest mõõdikutest on konversioonimäär: protsent külastajatest, kes sooritavad antud olukorda jõudes, näiteks veebilehte külastades, soovitud toimingu. Veebilehe laadimiskogemuse mõju konversioonimäärale on laialdaselt uuritud. Täheldatud mõju erineb aga uurimuste vahel märkimisväärselt, tingituna erinevatest laadimismõõdikutest ja uuritud populatsioonide erinevustest. Seetõttu tuleks usaldusväärsete tulemuste saamiseks mõõta mõju iga lehekülje jaoks eraldi. Antud töös käsitletakse probleemi, et TransferWise ei ole seda mõju mõõtnud ning seetõttu puudub ülevaade, kas lehe laadimiskogemust optimeerida või mitte, missuguses ulatuses ja missuguste mõõdikute osas.Selle taustal on käesoleval tööl kolm panust: (i) kasutaja tajutava laadimise mõõdikute defineerimine TransferWise'i kontekstis koos lahendusega nende mõõdikute kogumiseks TransferWise'i tootmiskeskkonnas; (ii) näidikulaua rakenduse implementatsioon laadimismõõdikute mõju konversioonimäärale mõõtmiseks; ja (iii) esialgne laadimismõõdikute ja konversioonimäära seose analüüs erinevate kanalite lõikes.Analüüsi tulemused kinnitavad oodatud korrelatsioone laadimisaja suurenemise ja konversioonimäära vähenemise vahel ning näitavad, et selle seose ulatus erineb laadimismõõdikute ja kanalite lõikes.
The user experience of loading a web page can have an effect on business metrics. One of these metrics is conversion rate: the percentage of visitors performing a desired action when reaching a given state such as visiting a web page. The effect of the page load experience on conversion rate has been widely studied. However, the observed effects vary significantly across studies due to the use of different load metrics and due to differences in the populations. Therefore, to ensure reliable results, this effect should be measured for each site separately. The problem addressed by this work concerns TransferWise not having measured this effect and therefore lacking insights on whether or not to optimise page load experience, to what extent, and with respect to which metrics.In this setting, the contributions of this work are threefold: (i) a precise definition of a set of user-perceived load metrics for TransferWise, together with a solution for collecting these metrics within TransferWise production environment; (ii) a dashboard application for measuring the effect of load metrics on conversion rate; and (iii) an initial analysis of the relations between load metrics and conversion rate for different acquisition channels.The results of the analysis confirm the expected correlations between increase in load time and decrease in conversion rate, and show that the extent of this relation varies by load metric and by acquisition channel.
The user experience of loading a web page can have an effect on business metrics. One of these metrics is conversion rate: the percentage of visitors performing a desired action when reaching a given state such as visiting a web page. The effect of the page load experience on conversion rate has been widely studied. However, the observed effects vary significantly across studies due to the use of different load metrics and due to differences in the populations. Therefore, to ensure reliable results, this effect should be measured for each site separately. The problem addressed by this work concerns TransferWise not having measured this effect and therefore lacking insights on whether or not to optimise page load experience, to what extent, and with respect to which metrics.In this setting, the contributions of this work are threefold: (i) a precise definition of a set of user-perceived load metrics for TransferWise, together with a solution for collecting these metrics within TransferWise production environment; (ii) a dashboard application for measuring the effect of load metrics on conversion rate; and (iii) an initial analysis of the relations between load metrics and conversion rate for different acquisition channels.The results of the analysis confirm the expected correlations between increase in load time and decrease in conversion rate, and show that the extent of this relation varies by load metric and by acquisition channel.