Platform Search Regulation and Sustainability: The case of Fashion E-Commerce
Type
conference contribution
Date Issued
2025
Author(s)
Abstract
Motivation. Sustainability has become a critical issue in contemporary commerce, particularly within the e-commerce sector, where consumers are increasingly concerned about the environmental impact of their purchases. In response, e-commerce platforms are introducing sustainability labels to help consumers make informed decisions about the environmental impact of their purchase decisions and to simplify the search for environmentally friendly products. Notable examples include Amazon’s Climate Pledge Friendly and Walmart’s recent announcement of green labeling for vendors. However, greenwashing—the deceptive promotion of products as environmentally friendly—has made it more difficult for consumers to make genuinely sustainable choices. Misleading environmental claims exacerbate this issue by obscuring the true sustainability of products, thereby increasing consumer confusion, mistrust and search costs. As a result, regulatory action against greenwashing has become a key priority for authorities worldwide, aiming to protect consumers from misleading environmental claims. Beyond issuing guidelines, regulators have directly intervened in multiple cases, including in Norway, the UK, and the US. These cases highlight the growing regulatory focus on improving transparency in sustainability claims. At the same time, there is ongoing debate about the most effective interventions to prevent greenwashing and ensure consumers have accurate information about a product's environmental standards. One approach is to provide more detailed sustainability information (e.g., "made of 30% recyclable materials" instead of "made of recyclable materials"), which can enhance transparency and discourage manufacturers from making misleading claims. However, consumers often face information overload when shopping online, making it unclear how effective increased transparency alone can be. An alternative approach is to implement stricter platform regulations, such as requiring verified sustainability claims for product recommendations or enabling filtering options based on verified sustainability criteria (e.g., recyclable materials). Despite these possibilities, there is limited research on the effectiveness of different policies in promoting sustainable consumption on digital platforms and avoiding greenwashing.
Case background and research questions. In this paper, we aim to address this gap by examining the impact of a regulatory decision that required a major European fashion platform to refine its sustainability information system and product recommendation/search functions. Among other requirements, the regulator mandated that the platform: First, update its search filters so that only products with a “substantial environmental promise” could be found. This criterion was defined as requiring products to meet a minimum sustainability threshold of 30%, as illustrated in Figure 1. In doing so, the filter option was temporarily removed completely from the platform. Second, enhance transparency regarding the magnitude of the environmental benefit. For example, as shown in Figure 2, products made from recycled materials must prominently display the exact percentage of recycled content. This unique setting allows us to address the following research questions:
1. To what extent do recommender systems for sustainable products influence consumer purchasing decisions in the fashion industry? Which types of brands experience the strongest impact?
2. Which strategy is more effective in reducing greenwashing: removing products with false sustainability claims from recommender (filtering) systems or enhancing transparency by providing detailed sustainability information?
3. What are the revenue implications of these two policies for greenwashed fashion products?
Challenge, method and data: A key challenge in measuring causal effects in our context is endogeneity in firm sustainability claims and consumer heterogeneity in purchasing behavior. To assess the impact of regulatory measures, we apply a Difference-in-Differences (DiD) research design. The treatment group consists of brands whose products were labeled as “sustainable” before the intervention, while the control group includes brands without sustainability labels. We ensure parallel trends by employing a synthetic control approach. Additionally, we assume SUTVA (Stable Unit Treatment Value Assumption) that there are no spillover effects between sustainable and unsustainable brands. We use daily product-level sales data from Grips Intelligence, a leading e-commerce insights provider (see https://gripsintelligence.com). Our dataset includes 15 days before and after the sustainability filter removal (Event 1) and the information update (Event 2), covering 420,000 SKUs of 3,700 brands across 24 fashion categories. Products are categorized by: a) Brand strength (strong vs. weak brands, according to Top 100 Best Global Brands of Interbrand), b) label type (binary labels vs. percentage-based sustainability indicators), and c) Sustainability degree (greenwashed which are <30% sustainable, 30-99% sustainable, or 100% sustainable).
We estimate two-way fixed effects models (unit and date fixed effects) with dependent variable being the natural logarithm of unit sales or revenue (incremented by one). The independent variables are those of a typical difference and difference specification with an indicator that equals 1 for our treatment group (TREAT, i.e., sustainable group) , else 0; an indicator that equals 1 for all days after the filter removal (PERIOD, including day of removal), and 0 for all days prior to the removal. We also control for the log(price) and the share of promotions on that day.
Preliminary Findings: Model-free evidence suggests that after removing the sustainability filter, sales of sustainable products declined more than non-sustainable ones, particularly for weak brands (Table 1). When additional sustainability information was introduced (Table 3), products with binary labels experienced a drop in sales, while those with continuous sustainability measures showed heterogeneous effects—with fully sustainable products benefiting from increased transparency. Our regression analysis confirms these trends:
1. Removing a product from the platform's sustainability recommendation filter reduces sales by -5.7% on average, while strong brands remained stable (see Table 2). This result suggest that advertising products based on its sustainability standards is more important for less known fashion brands.
2. Transparency measures negatively affected sales (-3.4%) for products that lost their green image due to stricter label recategorization but increased sales (+0.9%) for high sustainability products (i.e.., 100%), while greenwashed products don’t face significant changes (see Table4).
Implications and contribution. Our results indicate that consumers value sustainability characteristics and benefit from platform guidance, particularly when brand value is low. However, consumers respond more strongly to search visibility changes than to additional sustainability information. These findings align with research on digital platforms and choice architecture, which highlights how interface design influences user behavior and market competition (Molitor et al., 2020; Weinmann et al., 2016; Thaler, 2018). By bridging the literature on choice architecture, product sustainability, and consumer information, we contribute to the understanding of how search filters influence sustainable consumption. Our results build on prior studies on sustainability labeling (White et al., 2019; Tari & Trudel, 2023) and eco-friendly product positioning (Bollinger et al., 2020; Alé-Chilet & Moshary, 2022; Pachali et al., 2022; Araya et al., 2022; Kim et al., 2022; Adalja et al., 2022). This study provides empirical evidence that platform filtering nudges consumers toward highly sustainable products while discouraging those with minimal sustainability credentials, reinforcing the role of platforms in driving sustainable consumer behavior.
Case background and research questions. In this paper, we aim to address this gap by examining the impact of a regulatory decision that required a major European fashion platform to refine its sustainability information system and product recommendation/search functions. Among other requirements, the regulator mandated that the platform: First, update its search filters so that only products with a “substantial environmental promise” could be found. This criterion was defined as requiring products to meet a minimum sustainability threshold of 30%, as illustrated in Figure 1. In doing so, the filter option was temporarily removed completely from the platform. Second, enhance transparency regarding the magnitude of the environmental benefit. For example, as shown in Figure 2, products made from recycled materials must prominently display the exact percentage of recycled content. This unique setting allows us to address the following research questions:
1. To what extent do recommender systems for sustainable products influence consumer purchasing decisions in the fashion industry? Which types of brands experience the strongest impact?
2. Which strategy is more effective in reducing greenwashing: removing products with false sustainability claims from recommender (filtering) systems or enhancing transparency by providing detailed sustainability information?
3. What are the revenue implications of these two policies for greenwashed fashion products?
Challenge, method and data: A key challenge in measuring causal effects in our context is endogeneity in firm sustainability claims and consumer heterogeneity in purchasing behavior. To assess the impact of regulatory measures, we apply a Difference-in-Differences (DiD) research design. The treatment group consists of brands whose products were labeled as “sustainable” before the intervention, while the control group includes brands without sustainability labels. We ensure parallel trends by employing a synthetic control approach. Additionally, we assume SUTVA (Stable Unit Treatment Value Assumption) that there are no spillover effects between sustainable and unsustainable brands. We use daily product-level sales data from Grips Intelligence, a leading e-commerce insights provider (see https://gripsintelligence.com). Our dataset includes 15 days before and after the sustainability filter removal (Event 1) and the information update (Event 2), covering 420,000 SKUs of 3,700 brands across 24 fashion categories. Products are categorized by: a) Brand strength (strong vs. weak brands, according to Top 100 Best Global Brands of Interbrand), b) label type (binary labels vs. percentage-based sustainability indicators), and c) Sustainability degree (greenwashed which are <30% sustainable, 30-99% sustainable, or 100% sustainable).
We estimate two-way fixed effects models (unit and date fixed effects) with dependent variable being the natural logarithm of unit sales or revenue (incremented by one). The independent variables are those of a typical difference and difference specification with an indicator that equals 1 for our treatment group (TREAT, i.e., sustainable group) , else 0; an indicator that equals 1 for all days after the filter removal (PERIOD, including day of removal), and 0 for all days prior to the removal. We also control for the log(price) and the share of promotions on that day.
Preliminary Findings: Model-free evidence suggests that after removing the sustainability filter, sales of sustainable products declined more than non-sustainable ones, particularly for weak brands (Table 1). When additional sustainability information was introduced (Table 3), products with binary labels experienced a drop in sales, while those with continuous sustainability measures showed heterogeneous effects—with fully sustainable products benefiting from increased transparency. Our regression analysis confirms these trends:
1. Removing a product from the platform's sustainability recommendation filter reduces sales by -5.7% on average, while strong brands remained stable (see Table 2). This result suggest that advertising products based on its sustainability standards is more important for less known fashion brands.
2. Transparency measures negatively affected sales (-3.4%) for products that lost their green image due to stricter label recategorization but increased sales (+0.9%) for high sustainability products (i.e.., 100%), while greenwashed products don’t face significant changes (see Table4).
Implications and contribution. Our results indicate that consumers value sustainability characteristics and benefit from platform guidance, particularly when brand value is low. However, consumers respond more strongly to search visibility changes than to additional sustainability information. These findings align with research on digital platforms and choice architecture, which highlights how interface design influences user behavior and market competition (Molitor et al., 2020; Weinmann et al., 2016; Thaler, 2018). By bridging the literature on choice architecture, product sustainability, and consumer information, we contribute to the understanding of how search filters influence sustainable consumption. Our results build on prior studies on sustainability labeling (White et al., 2019; Tari & Trudel, 2023) and eco-friendly product positioning (Bollinger et al., 2020; Alé-Chilet & Moshary, 2022; Pachali et al., 2022; Araya et al., 2022; Kim et al., 2022; Adalja et al., 2022). This study provides empirical evidence that platform filtering nudges consumers toward highly sustainable products while discouraging those with minimal sustainability credentials, reinforcing the role of platforms in driving sustainable consumer behavior.
Language
English (United States)
Keywords
platform search regulation
greenwashing
fashion e-commerce
HSG Classification
contribution to scientific community
Event Title
Statistical Challenges in Electronic Commerce Research (SCECR)
Event Location
Paphos
Event Date
24. June 2025
Subject(s)
Division(s)