[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85279-en":3,"doc-seo-85279-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85279,1374391974585,"Genevieve","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","Dimensionality in Satisfaction Ratings","Validating LLM annotation as a way to recover decomposed customer-care satisfaction across five experience axes: overall, agent, outcome, product, and customer effort. A GPT-4.1 model annotated ~9,000 support conversations from a global consumer-goods firm, and annotations were validated against customer self-reported ratings. Four axes track closely, while product satisfaction shows weak correspondence. Misalignments cluster in a readable divergent tail, not general drift; correlations rise to 0.811 and 0.914 after excluding severe divergences. Axes are highly collinear, so decomposition supports attribution and coverage rather than incremental prediction, and census-level satisfaction is lower than survey reports.","Dimensionality in Satisfaction Ratings  \nValidating LLMannotation as a method to recover decomposed satisfaction across five axes of customer experience.  \nAndrew Hong Jason Potteiger  \nDimension Labs  \nJuly 2026  \nAbstract  \nWe used a large language model (GPT-4.1) to annotate the text of ~9,000 support conversations at a global consumer-goods firm, decomposing customer-care satisfaction into component axes (overall, agent, outcome, product, and customer effort), and validated the LLM annotations against the satisfaction ratings customers gave themselves. Four of five axes track self-reported satisfaction closely (overall, agent, and outcome near an unadjusted 0.65; effort −0.54), while product satisfaction is weak against the available proxy. The unadjusted correlation also understates the alignment: the disagreements concentrate in a small, readable tail of divergent sessions rather than in general drift, and the overall correlation rises to 0.811 when only the severe divergences are excluded and to 0.914 when the full divergent tail is excluded. The axes are also highly collinear, and adding them to the overall score does not improve prediction of the customer’s rating: the decomposition’s value is not incremental prediction but attribution and coverage. And, with greater coverage the picture of the data changes. Read on every contact rather than the few that return a survey, satisfaction is markedly lower than the survey reports (a fullcensus 2.91 against the surveyed 3.62 on a five-point scale) . The promise of decomposed satisfaction as a methodology is the ability to identify more nuanced drivers of customer experience in conversational data.  \nFindings at a glance. Each row names the cohort behind the number; sections give the full evidence.  \nFinding Headline number Cohort Section  \nThe annotation tracks the customer’s own rating  \nThe component axes add no predictive power  \nThe survey overstates satisfaction  \nThe misses are readable, not random  \nρ = 0.65 unadjusted; 0.811 / 0.914 with  \nsevere / all divergences excluded  \nInter-axis ρ of 0.84–0.94; composites land within 0.02 of the overall axis alone  \nSurveyed sessions average 3.62; the full census reads 2.91  \n60 of 83 divergent sessions have a  \npredictable pattern with identifiable frame  \nIn-chat base (n = 385)  \nCollinearity sample (3,310)  \nThe census (8,880)  \nDivergent set (83)  \n§§4, 7  \n§5  \n§6  \n§7  \n1. Introduction  \nEvery customer-care organization faces the same asymmetry between what it hears and what it measures. The global consumer-goods firm in this study handled 8,880 support conversations across an automated chat assistant and recorded voice calls. Only 537 of them, about six percent, ended with a satisfaction rating the customer chose to give; the other 8,343 left a transcript and no ratings. Satisfaction programs are built on the rated minority, treated as the voice of the customer, even though the customers who supply the ratings select themselves into the measurement (there have been few alternatives) . A single overall rating is also a compression: it reduces an interaction involving an agent, an outcome, a product, and some amount of customer effort to one holistic verdict. This paper asks whether a large language model, applied as an annotation instrument, can read the transcript and recover the parts an overall rating compresses, producing a satisfaction score for each component of the experience. It then asks whether, once those annotations are validated against what customers say about themselves, reading them on every contact rather than the selfselected few reveals something the survey cannot.  \nThe annotations were produced by a single large language model (GPT-4.1) applied as a recognition instrument: it read one conversation’s text at a time and returned a structured set of satisfaction scores for that conversation alone. For each session the model assigned an overall satisfaction score and four component scores (satisfaction ","cbCair2md1zZKawY","https://ap.wps.com/l/cbCair2md1zZKawY","pdf",1064811,1,25,"English","en",105,"# Abstract\n# Findings at a glance\n# Introduction\n## Study populations and validation approach","[{\"question\":\"How did the study use GPT-4.1 to measure customer satisfaction dimensions?\",\"answer\":\"The model read each support conversation and output a structured set of satisfaction scores: one overall score and four component scores covering agent, outcome, product, and customer effort.\"},{\"question\":\"Which satisfaction axes matched customer self-reported ratings most closely?\",\"answer\":\"Overall, agent, and outcome tracked self-reported satisfaction closely, with correlations near 0.65 unadjusted, while effort showed a weaker but consistent relationship and product satisfaction was weak against the available proxy.\"},{\"question\":\"Why do the authors say the product axis is weak and how do divergences affect reported correlations?\",\"answer\":\"Disagreements concentrate in a small divergent tail rather than general drift. Excluding severe divergences increases the correlation to 0.811, and excluding the full divergent tail raises it to 0.914.\"}]",1784202230,63,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"dimensionality-in-satisfaction-ratings","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/dimensionality-in-satisfaction-ratings/85279/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"How did the study use GPT-4.1 to measure customer satisfaction dimensions?","Question",{"text":75,"@type":76},"The model read each support conversation and output a structured set of satisfaction scores: one overall score and four component scores covering agent, outcome, product, and customer effort.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"Which satisfaction axes matched customer self-reported ratings most closely?",{"text":80,"@type":76},"Overall, agent, and outcome tracked self-reported satisfaction closely, with correlations near 0.65 unadjusted, while effort showed a weaker but consistent relationship and product satisfaction was weak against the available proxy.",{"name":82,"@type":73,"acceptedAnswer":83},"Why do the authors say the product axis is weak and how do divergences affect reported correlations?",{"text":84,"@type":76},"Disagreements concentrate in a small divergent tail rather than general drift. Excluding severe divergences increases the correlation to 0.811, and excluding the full divergent tail raises it to 0.914.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]