Authors:
Alexander Seewald
1
;
Thomas Wernbacher
2
;
3
;
Mario Platzer
4
and
Alexander Pfeiffer
2
Affiliations:
1
Seewald Solutions GmbH, Lärchenstraße 1, 4616 Weißkirchen a.d. Traun, Austria
;
2
Universität für Weiterbildung Krems, Dr.-Karl-Dorrek-Straße 30, 3500 Krems, Austria
;
3
Liberacerta e.U., Am Lindenhof 37/11, 8043 Graz, Austria
;
4
yVerkehrsplanung GmbH, Brockmanngasse 55, 8010 Graz, Austria
Keyword(s):
AI for Green, Characterizing Returns, Fitting Tool, Depth Cameras.
Abstract:
Within the context of the Green eCommerce project where we build tailored add-ons for webshops to increase climate-friendly shipping, we analyzed reasons for returns using a modified rule learning algorithm but found no actionable rules. However, since many returns are driven by wrong size information, we have also developed a prototype Fitting Tool app that uses active depth sensing to measure several relevant body measurements and uses these to estimate T-Shirt sizes. Although these body measurements could be shown to be quite precise, T-Shirt sizes could only be predicted at low accuracy. On the other hand, self-reporting by test users showed that the perceived accuracy was considered 1.5-3x higher. Analyzing this issue, it was found that the reason for this is most likely manufacturer bias in reported size, which will be addressed in future work.