Glossier engages customers on numerous data-rich platforms. \u00a0Customers visit the brand\u2019s originating blog (Into the Gloss<\/em>), browse Glossier\u2019s website, interact with Glossier\u2019s social media, and visit physical stores [2]. \u00a0Customer-brand interactions provide data that informs predictions of when behaviors will turn profitable.\u00a0 Glossier\u2019s blog and social media provide platforms for customers to voice opinions on Glossier products and desires for new ones.\u00a0 This offers opportunity for Glossier to predict what products or features will perform well in the future, a process currently done manually but which machine learning could automate [3].<\/p>\n Glossier\u2019s massive amounts of data from varied sources result in data that is both \u201cbig\u201d and \u201cwide.\u201d\u00a0 As Glossier grows, its data becomes bigger and wider.\u00a0 Glossier must distinguish \u201cout-of-sample\u201d from \u201cout-of-context\u201d data before mistakenly applying a model informed by a subset of customer interactions (e.g. online) to a different category of interactions (e.g. in-person). [4]\u00a0 Further, Glossier must use caution when automating the digestion of customer-brand engagement if these interactions are core to its identity.<\/p>\n Short Term<\/u><\/strong><\/p>\n Glossier recognized the need to connect disparate data upon wrongly assuming blog readers would become online customers [3].\u00a0 Though the brand\u2019s launch spurred blog traffic (see Figure 1), existing readers did not equate to revenues on Glossier.com. \u00a0To study customer behavior fluidly between online platforms, Glossier hired marketing tech firm Segment [5].\u00a0 Using cross-domain machine learning, Glossier tracked identities across the channels to identify profitable patterns.\u00a0 Glossier discovered that blog readers read mobile articles, followed links to shop on Glossier.com, then completed transactions via desktop.\u00a0 Now, Glossier recognizes this pattern and proactively provides customers with desktop and mobile links between the blog and website. [3]\u00a0 Glossier also sources data from email, social media, and physical stores.\u00a0 Segment helps store this data and feeds it to machine learning algorithms [5].<\/p>\n Glossier relies on customer engagement to inform product development, once using a blog post to crowdsource its facial cleanser development [6].\u00a0 Glossier manually digests \u201ccomments left on articles, social media, user-generated content posts and product pages,\u201d but plans automate this process with machine learning [3].<\/p>\n Figure 1<\/u><\/strong><\/p>\n Source: Data excerpted from We Analyzed 9 Of The Biggest Direct-to-Consumer Success Stories To Figure Out The Secrets to Their Growth \u2014 Here\u2019s What We Learned<\/em> (CBInsights, 2017).<\/p>\n Further Ahead<\/u><\/strong><\/p>\n CEO Emily Weiss is planning Glossier\u2019s \u201cPhase Two,\u201d during which the brand will create a \u201csocial-selling\u201d website providing customers with a \u201csocial media and shopping mashup\u201d [2].\u00a0 A social commerce development will provide richer data from which to learn customer behavior patterns and desires.\u00a0 With these plans and 37% of its employees deemed \u201ctechnologists\u201d, Glossier dubs itself a technology company [1].<\/p>\n Recommendations<\/u><\/strong><\/p>\n A social-selling app could close gaps between data sources, just as Glossier did across online channels.\u00a0 Glossier could incorporate member profiles into blog browsing, online shopping, and in-person shopping with member check-ins or location recognition.\u00a0 With cross-channel tracking of customer behavior, machine learning could identify profitable patterns and tailor customer experiences accordingly.\u00a0 For example, a common purchase path might involve a customer viewing a product online before purchasing in person when geographically feasible.\u00a0 Glossier could identify the start of this pattern online and prompt purchases by inviting customers to visit nearby stores.<\/p>\n Glossier\u2019s data sources are growing at a faster rate than what can be developed to learn from them.\u00a0 Figure 2 plots traffic on Glossier.com since its 2014 launch.\u00a0 Machine learning only accurately answers isolated questions such that \u201cthe data you feed to your learning algorithm includes [all] there is to the problem\u201d [7].\u00a0 For example, Glossier identified customer behavior patterns online that generated purchases.\u00a0 New physical stores might significantly alter the purchase journey.\u00a0 Until data from in-person interactions is incorporated into Glossier\u2019s learning algorithms as recommended above, the algorithm that learned solely from online behavior will capture a limited picture.<\/p>\n Utilizing Glossier\u2019s wealth of data warrants caution.\u00a0 Data-rich organizations often derive findings beyond the scope of KPIs.\u00a0 Using machine learning, Glossier must cut through massive amounts of data to discern which performance indicators are key [8].<\/p>\n Figure 2<\/u><\/strong><\/p>\n Source: Data excerpted from We Analyzed 9 Of The Biggest Direct-to-Consumer Success Stories To Figure Out The Secrets to Their Growth \u2014 Here\u2019s What We Learned<\/em> (CBInsights, 2017).<\/p>\n Open Questions<\/u><\/strong><\/p>\n Looking ahead at Glossier\u2019s employment of machine learning, questions remain:<\/p>\n Although a social-selling app provides promise, will it be feasible for Glossier to analyze customer behavior across all channels seamlessly, or will disparate data need to be learned from as such, limiting machine learnings to the extent of customer behavior in individual channels?<\/p>\n Glossier plans to use machine learning to analyze customer engagement to inform product development, a process currently done when the brand interacts with customers\u2019 comments and crowdsourced ideas online.\u00a0 Will this automation hinder customers\u2019 personal connections with Glossier and tarnish Glossier\u2019s brand?<\/p>\n <\/p>\n (798 words)<\/p>\n <\/p>\n \u00a0<\/strong><\/p>\n \u00a0<\/strong><\/p>\n \u00a0<\/strong><\/p>\n \u00a0<\/strong><\/p>\n \u00a0<\/strong><\/p>\n \u00a0<\/strong><\/p>\n \u00a0<\/strong><\/p>\n \u00a0<\/strong><\/p>\n <\/p>\n","protected":false},"excerpt":{"rendered":" Glossier engages its customers on numerous platforms, each providing this beauty brand with a wealth of data. Glossier is feeding this data into machine learning algorithms to better direct the customer down a purchasing path and to automate the crowdsourcing of product development.<\/p>\n","protected":false},"author":11875,"featured_media":35649,"comment_status":"open","ping_status":"closed","template":"","categories":[186,298,2122,219,346,16],"class_list":["post-35515","hck-submission","type-hck-submission","status-publish","has-post-thumbnail","hentry","category-beauty","category-big-data","category-data-analytics","category-ecommerce","category-machine-learning","category-retail","hck-taxonomy-organization-glossier","hck-taxonomy-industry-beauty-and-cosmetics","hck-taxonomy-country-united-states"],"connected_submission_link":"https:\/\/d3.harvard.edu\/platform-rctom\/assignment\/rc-tom-challenge-2018\/","yoast_head":"\n![\"\"](\"https:\/\/d3.harvard.edu\/platform-rctom\/wp-content\/uploads\/sites\/4\/2018\/11\/glossier-1.png\")
<\/a><\/p>\n<\/a><\/p>\n\n