Returns are part of everyday life in online retail. Especially in the fashion segment, they pose a strategic problem—with significant economic and environmental consequences. At the same time, studies and industry reports show that traditional measures are reaching their limits, as return rates continue to rise. The key increasingly lies in a better, data-driven understanding of returns and in the ability to anticipate and, ideally, prevent them before they occur. In practice, “anticipation” can refer to different areas: improved data collection, process and IT standardization, AI-supported decisions at specific points along the customer journey, reverse logistics, or even returns forecasting can all play a role.
Why returns are a major issue in fashion retail
Returns are not a marginal phenomenon, but a structural feature of German online retail. There are no official statistics, but for 2025 returns experts and industry associations estimate that more than 550 million parcels were sent back in Germany. This magnitude also comes in the context of a growing online market, suggesting that the problem could become even more severe in the future.
When it comes to returns, the fashion sector stands out in particular. According to a customer survey, clothing and shoes – together accounting for 47.5% – are among the most frequently returned items in e-commerce. According to the Returns Management research group, around 90 % of returned items overall come from the fashion segment. Studies also show that a large share of fashion retailers report overall return rates of up to 50 %. This figure is significantly lower in other product categories.
In comparison with other European countries, Germany also plays a questionable special role , as returns are particularly prevalent here and Germany is considered the „European returns champion“. Factors such as the high share of purchases on invoice, generous return policies, and typically free return shipping encourage consumer behavior in which sending items back has almost become the norm and the home is increasingly becoming the new fitting room.
The impact is substantial: according to EHI, shipping, inspection, reconditioning or write-offs cause average costs of five to ten euros per returned item. In addition to burdens on logistics systems, the growing environmental impact of returns through CO₂ emissions is a growing problem and is putting increasing pressure on retailers’ sustainability goals.
What causes returns in fashion e-commerce?
Returns rarely happen by chance, and research shows recurring causes. In addition to quality issues and damage, the following drivers are of particular interest for returns management:
Size and fit issues
Consumer surveys identify uncertainty around size and fit as the main reasons for returns: in a KPMG/EHI consumer survey, “wrong size” is cited by 72.6% as a trigger for returns. Online shoppers do not have products “physically in front of them” and often find it difficult to assess fit information.
Inaccurate or insufficient product descriptions
According to the KPMG Consumer Snapshots , inaccurate or insufficient product descriptions, or deviations of the product from product images are a driver of returns. Discrepancies between how an item is presented in the shop and the actual product create a mismatch between customer expectations and reality.
Multiple orders for selection
How do customers deal with these difficulties? They order multiple versions of a product. Industry experts point out that deliberate multiple orders of several sizes or variants with the clear intention of sending most of them back is a common behavior.
Measures to reduce returns
Measures in the returns context can be grouped into four major areas. Since not every retailer starts at the same maturity level, it makes sense to view the building blocks as combinable. In practice, one thing becomes clear: impact comes less from individual tools and more from the clean integration of data, processes, decision logic, and AI.
1) Foundations: Establish a data basis
A practical reference point: according to EHI, 81% of retailers systematically capture reasons for returns; 67% derive concrete measures from them. That sounds like a lot, but it also means that nearly 20% of retailers do not record return reasons at all. Surprising, but explainable. The prerequisite for systematically analyzing returns and deriving suitable countermeasures is having all relevant information available in full. This may require digitizing processes first, or standardizing and making usable existing data that may already exist in a distributed, disconnected form. In practice, data often needs standardization to be usable (return-reason codes), item/variant data (e.g., size/cut) must be properly maintained, and there must be a reliable linkage between shop, PIM, OMS, WMS, and possibly CRM data. Without these foundations, both process automation and more complex AI approaches remain limited in their impact.
Read our articles on data harmonization: „Reconciliation – Turning Data Chaos into Clarity”, From Mapping to Blending – Clarifying Data Integration Terminology and Navigating the Challenges of Data Reconciliation: An Example Use-Case.
2) Prevention: Reduce returns before they occur
Improved product presentation to lower the return rate
Many fashion retailers are already trying to reduce returns. According to an EHI survey, retailers often rely on measures such as more detailed product information and images to reduce returns. These are intended to help customers make better purchasing decisions so that the items they buy are kept.
Personalisation
As we already know, especially for clothing and shoes, size and fit issues are among the main drivers of returns. This root-cause area can be addressed through improved personalization in online retail. In addition to classic recommendations such as size charts and transparent customer reviews, interactive tools are also intended to help users choose the right product. Fit and size recommendations have a measurable impact on returns, especially when they use customer-specific data. Nevertheless, customers still cite difficulties in this area as one of the main reasons for their returns.
Our article on personalization in online fashion retail can be found here.
Rules and incentives
Both positive and negative incentives are common tools in the fight against returns. Positive incentives include, for example, discounts or vouchers for keeping the goods. Examples of negative incentives are fees, return shipping costs borne by the buyer, restrictions on payment options, or limits on basket value. A study by IFH KÖLN and AZ Direct describes that positive incentives show better effects than negative ones—however, it also notes that the effect still depends on the customer’s “return type,” and that more than 23% of surveyed customers could not be deterred from returning items by any of the measures. Positive and negative incentives therefore have an effect, but not always to the hoped-for extent. And according to a Statista survey, paid returns could also deter roughly half of customers from making a purchase.
Combinations of positive and negative incentives are also used: for example, Asos introduced targeted treatment of customer groups with different return behavior. On the one hand, customers are given a transparent view of their return rate. On the other hand, groups with particularly high rates are charged return costs. At the same time, they are shown what behavior can improve the rate in order to avoid costs in the future.
3) Processing: Make returns processes efficient
In addition to prevention, efficiency in returns logistics is a lever in its own right: even if the return rate hardly decreases in the short term due to changes in this area, cycle times, process costs, and the ability to resell returned goods are key factors in determining the resulting economic damage.
Typical measures include, for example:
- Digitized returns workflows (self-service portal, automated status communication)
- Standardized status models and workflows (announced → in transit → received → inspected/graded → resellable → restocking/reconditioning/write-off)
- Automation/partial automation of repetitive steps (e.g., scans, assignment, rule sets for exceptions)
4) Management & Analytics: From transparency to forecasting
Returns and AI
The basic measures discussed so far are important and form part of a solid baseline strategy. They are most effective where they directly improve buyers’ information base or reduce operational errors (e.g., better descriptions). However, the effectiveness of traditional methods is limited. Many retailers also lack a strategic integration of these measures, so their efforts address only individual causes. Basic measures alone are often not sufficient to sustainably capture the complex mechanisms behind high return rates—additional, data-driven forecasting and control tools are needed that identify causes in early purchasing phases and can intervene proactively. An EHI survey shows: only 7.3% of retailers already use AI in returns management, but 45.5% consider it relevant in the future.
Retrospective analysis
According to EHI, 81% of retailers capture reasons for returns. Such analyses are professionally necessary and form the basis for fundamental optimization measures, such as improving product descriptions or assortment decisions. Nevertheless, return rates continue to rise, even though EHI also reports that 67% of retailers already derive concrete measures from recorded return reasons. However, the approaches taken often remain reactive: returns only become visible after they have occurred. New products, new target groups, or changed contexts (e.g., price promotions, weather, trends) can only be anticipated to a limited extent with purely descriptive methods. Especially in fashion retail, a sector with strong seasonality and product novelty, future events are difficult to predict based solely on historical data—an effect also referred to as the „cold start problem which for example also affects product recommendations and demand forecasting in fashion retail.
Forecasting approaches: Thinking of returns as probabilities
Returns forecasting refers to the data-driven prediction of return probabilities or volumes before purchase or shipment takes place. Such forecasts can relate to individual products or categories, customer segments with similar behavior, or the specific combinations of customer, product, and context.
Research from MIT Sloan and other groups shows that predictive models in fashion retail can help identify products with a high return risk. By no longer showing these products to customers, a profit increase could be achieved.
How returns forecasting is implemented technically
Methodologically, both classic machine learning methods and modern AI approaches are used. Classic machine learning methods (e.g., logistic regression, random forests, gradient boosting) are comparatively easy to explain, robust, and particularly well suited to structured data such as prices, sizes, categories, or historical return rates. Modern AI models, for example those based on Graph Neural Networks , enable a more fine-grained representation of complex interactions between customer, product, and interaction data that is often not achievable with traditional approaches.
There is therefore no “one-size-fits-all” AI approach for return prediction. Solutions must be selected to fit the specific data situation and the task at hand. Regardless of the specific method, however, one thing holds true: the technical and mathematical implementation of generating return forecasts is only part of the job. At least as important are topics such as data integration, governance, validation, and embedding the forecasts into operational decision-making.
Predictive approaches such as returns forecasting do not replace traditional measures. Rather, they help to reorganize them: instead of optimizing broadly and across the board, it becomes possible to allocate effort in a targeted way where a high return risk is predicted.
Retailers can then strategically tackle and make forward-looking use of aspects such as:
· prioritized maintenance of product content or size information
· assortment and purchasing decisions (including as early as the product design process)
· targeted management of campaigns and price promotions
· personalized notices or alternative suggestions before checkout
· early logistics capacity planning
This fundamentally shifts how returns are handled: from retrospective correction to proactive management.
Why Returns Forecasting rarely works “off the shelf”
At first glance, returns forecasting may look like a classic use case for standardized analytics or AI products. In practice, however, it becomes clear that generic plug-and-play solutions in fashion retail often provide only limited predictive value.
Returns arise from a complex interaction of product characteristics, customer behavior, situational factors, and company-specific data landscapes and processes. Which influencing variables are relevant—and to what extent—differs significantly across retailers, assortments, price segments, and target groups. A model that does not explicitly account for these differences inevitably remains coarse and loses relevance for operational decision-making. Returns forecasting is less a ready-made solution than an individualized design process at the intersection of data, processes, and business logic. Standard building blocks can help, but lasting value is created when analysis, domain expertise, and technical implementation are deliberately aligned.
Conclusion
Returns in fashion e-commerce are not an isolated problem, but arise from the interplay of product data, customer expectations, purchasing behavior, and operational processes. Accordingly, measures work best as a modular toolkit that can be combined: first, retailers create transparency by establishing solid data and system foundations; building on that, preventive measures during the purchasing phase and efficiency levers in reverse logistics can be implemented. Analytics – and, where appropriate, returns forecasting – then help prioritize limited resources in a targeted way. Which steps make sense to take first depends on the maturity of the data, processes, and organization.
At CID, we have over 25 years of experience developing tailored e-commerce solutions for the retail sector. We focus on long-term partnerships and solutions that are individually aligned with each customer’s business, data, and technology ecosystem. Get in touch – we’ll help you assess your situation and identify, develop, and implement impactful use cases.
