How Otrium deployed Autostore robotics and AI for warehouse efficiency
The off-price platform automated its fulfilment centre with Autostore's grid system and generative AI for product imagery, cutting throughput time and accelerating sell-through.

KEY TAKEAWAYS Summary by the editors
- Otrium implemented Autostore's robotic grid system in its European distribution centre in 2025 to handle higher order volumes.
- The platform uses generative AI to create product imagery when brand assets are unavailable, reducing time to list inventory.
- Automated storage and retrieval systems enable fashion retailers to process returns and new stock faster than manual warehouses.
- Combining warehouse automation with AI-generated content can shorten the markdown cycle and improve sell-through velocity for off-price inventory.
Otrium, the Amsterdam-based off-price fashion platform, completed the installation of an Autostore automated storage and retrieval system at its European fulfilment centre in early 2025, according to company announcements and logistics industry sources. The deployment came as the retailer sought to handle growing order volumes and reduce the time between receiving inventory from brand partners and shipping to customers across Europe. At the same time, Otrium integrated generative AI tools to produce product photography when brand-supplied images were missing or unsuitable, cutting the delay between stock arrival and online listing.
The dual investment in warehouse robotics and content automation addresses a core challenge in off-price retail: inventory moves faster when items are photographed, listed and picked quickly, yet many end-of-season or overstock garments arrive without digital assets or in packaging incompatible with high-speed manual sortation. For fashion executives weighing automation spending, Otrium's case illustrates how robotics and machine vision can work together to compress the markdown cycle and recover margin from unsold inventory.
This article examines Otrium's 2025 warehouse and content automation projects, the throughput and velocity gains the company reported, and the broader implications for fashion supply chains managing excess stock and returns.
What is Otrium and why did it automate its warehouse?
Otrium operates an online marketplace that sells past-season and overstock inventory directly from fashion brands, working with more than 400 labels including Patagonia, Karl Lagerfeld and Scotch & Soda. The platform takes possession of unsold goods, photographs and lists them on its site, and ships orders from its own fulfilment centres. Revenue grew as brands sought alternatives to heavy discounting in their own channels, and order volumes in Europe increased sufficiently by 2024 that manual picking and packing became a bottleneck, according to supply chain industry observers.
In late 2024, Otrium announced plans to install Autostore, a Norwegian provider of cube-based automated storage systems, at its main European distribution centre. The system uses a grid of stacked bins and autonomous robots that retrieve containers on demand, then deliver them to picking stations where human workers pack orders. Autostore systems can achieve storage densities up to four times higher than conventional shelving and reduce walking time for warehouse staff by bringing bins to fixed workstations. The installation went live in the first quarter of 2025, enabling Otrium to process a higher volume of small, mixed SKU orders typical of online fashion retail.
How does Autostore's robotic grid system work in a fashion warehouse?
An Autostore grid comprises aluminium tracks arranged above a storage area, with robots travelling on top to lift and move bins stored in vertical columns below. When an order arrives, warehouse management software dispatches a robot to the grid location holding the required item. The robot lifts the bin, transports it to a port at the edge of the grid, and a human operator picks the garment, scans the barcode, and places it in a shipping carton. Empty bins return to the grid, and the system prioritises frequently picked items in upper layers to minimise retrieval time.
For fashion inventory, the approach suits high-mix, low-volume picking: each bin may contain only a few units of a single size and colour, and orders often include items from multiple brands and categories. Autostore's compact footprint also allows higher inventory density than traditional racking, an advantage in urban distribution centres where rent is high. The robots themselves require no special flooring or guidance infrastructure, and the modular design means capacity can expand by adding more grid modules or robots without redesigning the facility.
What results has Otrium reported from warehouse automation?
Otrium has not published detailed throughput metrics in investor filings, as the company remains privately held. However, industry sources and logistics analysts note that Autostore installations in e-commerce typically double or triple picking rates compared to manual operations, measured in units per hour per worker. In 2025, automated storage and retrieval systems in fashion warehouses achieved average order cycle times of under 30 minutes from order receipt to dispatch, compared to 60 to 90 minutes in manual facilities of similar scale. Otrium's management stated in a 2025 trade interview that the system allowed the company to handle seasonal peaks without hiring additional temporary staff, suggesting labour productivity gains in line with industry benchmarks.
Equally important for an off-price model is the speed of inbound processing. When a brand ships excess inventory to Otrium, every day the goods spend in receiving, photography and listing represents lost selling time and increased risk of further markdowns. Autostore's bin-based storage means items can be putaway immediately after photography and quality checks, without waiting for a free shelf location or forklift availability, compressing the time from delivery to availability for sale.
Source: Supply Chain Dive, 2025
How is Otrium using generative AI for product imagery?
Many garments that arrive at Otrium lack brand-supplied photographs or have images that do not meet the platform's style standards: consistent lighting, plain backgrounds and multiple angles to reduce returns. Historically, Otrium re-photographed items in-house, adding cost and delay. In mid-2025, the company began testing generative AI tools that create product images from a smartphone snapshot or even text descriptions of the garment, according to retail technology reports.
Generative AI models trained on fashion product photography can produce studio-quality images by transforming casual photos into standardised views with controlled lighting and background removal. Otrium's implementation allows warehouse staff to capture a quick image on a mobile device; the AI system then generates a set of listable photos in seconds, which are uploaded to the product page. The company reported that this workflow reduced the median time to list a new SKU by approximately 40 per cent compared to manual studio photography, though it did not disclose absolute time figures.
Beyond speed, AI-generated imagery can improve consistency across thousands of SKUs from different brands, each with varying original photography quality. For customers browsing a mixed assortment, uniform image style reduces cognitive friction and may lower return rates by setting accurate expectations for fit and colour. Otrium's data, shared in a 2025 industry presentation, indicated that listings with AI-standardised images saw return rates 10 to 15 per cent lower than those using unedited brand photos, though the company cautioned that multiple variables influence returns and the effect varied by category.
Which other fashion companies are automating warehouses and content?
ASOS installed Autostore systems at its Lichfield and Barnsley fulfilment centres in the United Kingdom between 2022 and 2024, expanding robotic capacity to handle over 100,000 items per day across both sites. The online retailer cited the technology as central to managing peak demand without proportional increases in labour costs. ASOS also deployed AI tools for product tagging and image enhancement in 2024, automating the creation of metadata and cropping thousands of product photos daily.
In 2023, Zalando integrated automated guided vehicles and robotic picking arms at its fulfilment centre in Lahr, Germany, aiming to increase parcel throughput by 30 per cent. The pan-European platform has separately invested in virtual try-on and AI-generated size recommendations, linking warehouse efficiency gains to front-end tools that reduce returns and improve inventory turn. Zalando reported in its 2024 annual report that logistics cost as a percentage of revenue fell by 1.2 percentage points year-on-year, attributing part of the improvement to automation investments.
Smaller and mid-market players are following similar paths. About You, the Hamburg-based online fashion retailer, announced in early 2025 that it would deploy AI-based product image generation across its assortment, targeting a 50 per cent reduction in photography lead time for third-party brands selling on its marketplace. The move aimed to shorten the gap between a brand uploading inventory data and the item appearing live on the site, directly influencing sell-through velocity for time-sensitive fashion goods.
- 2022ASOS begins Autostore rollout at Lichfield fulfilment centre, UK
- 2023Zalando integrates automated guided vehicles and robotic arms at Lahr, Germany
- 2024ASOS deploys AI for product tagging and image enhancement
- Late 2024Otrium announces Autostore installation at European distribution centre
- Q1 2025Otrium Autostore system goes live; order processing capacity increases
- Mid-2025Otrium pilots generative AI for product imagery, cutting listing time by 40%
- Early 2025About You commits to AI image generation for marketplace sellers
Source: ASOS, Zalando, Retail Dive, FashionUnited
What are the return on investment and payback periods for warehouse robotics?
Capital expenditure for an Autostore system depends on grid size, number of robots and integration complexity. Industry estimates place the cost of a mid-scale installation, suitable for processing 20,000 to 40,000 orders per day, at between 5 million and 10 million euros, including hardware, software and commissioning. Autostore itself reported in 2024 that typical customer payback periods ranged from three to five years, driven by labour savings, increased throughput and reduced real estate costs per cubic metre of storage.
For fashion e-commerce, the ROI calculation must account for seasonal peaks and the cost of temporary labour. A manual warehouse may double headcount during the November-to-January period; an automated facility can handle the same volume by running robots on additional shifts without hiring or training surge staff. Labour arbitrage is less pronounced in Western Europe than in some markets, but warehouse wages in the Netherlands, Germany and the United Kingdom have risen 15 to 25 per cent since 2021, according to Eurostat data, making automation more attractive even at current interest rates.
Energy consumption is another factor. Autostore robots are battery-powered and comparatively light, drawing less power than conveyor-based systems. The company states that its systems consume approximately 0.1 kilowatt-hours per bin retrieval, translating to daily energy use of around 100 kilowatt-hours for a 20-robot grid handling 10,000 picks. In a facility previously reliant on forklifts and extensive lighting across wide picking aisles, the compact grid and reduced floor area can lower electricity and heating costs, contributing to overall ROI.
How does automation affect markdown velocity and sell-through?
Markdown velocity refers to the speed at which discounted or off-price inventory sells through. For Otrium, faster processing from inbound to listing means garments reach customers while still seasonally relevant, commanding higher residual prices and reducing the depth of subsequent markdowns. If a spring jacket arrives in March but spends three weeks in photography and putaway, it may not list until late April, requiring a steeper discount to clear before summer. Cutting that delay to under a week preserves pricing power.
Warehouse automation contributes by ensuring inventory is available for immediate dispatch once an order is placed. Stockouts due to mislaid bins or picking errors force customers to substitute or cancel, harming conversion and extending the tail of slow-moving SKUs. Automated systems typically achieve inventory accuracy rates above 99.5 per cent, compared to 95 to 98 per cent in manual operations, reducing phantom stock and improving on-shelf availability. Higher accuracy means pricing algorithms can rely on real-time stock counts, optimising discounts to clear inventory without over-discounting high-demand items.
Generative AI for imagery acts as a parallel accelerator. When a sought-after item can be listed the same day it arrives, the platform captures demand that might otherwise go to a competitor. The combined effect of rapid putaway, accurate picking and instant content creation shortens the cash conversion cycle: brands receive payment sooner, and Otrium turns inventory faster, freeing working capital for the next batch of goods.
What are the risks, limitations and open questions?
Warehouse robotics require upfront capital and ongoing maintenance contracts, creating fixed costs that can burden operators if order volumes fall. Fashion e-commerce has seen consolidation and several high-profile insolvencies since 2022; a company that installs expensive automation and then experiences a demand shock may face higher operating leverage and cash burn than a manual competitor able to flex labour quickly. Otrium's private ownership means limited public disclosure of cash flow and debt, making it difficult for outside observers to assess the financial resilience of its automation bet.
Generative AI for product imagery also carries risks. While the technology has improved, it can still produce artefacts or colour inaccuracies that increase return rates if customers receive items that do not match the AI-generated photo. Consumer protection regulators in the European Union have signalled that synthetic media in e-commerce must be clearly labelled, and brands may object to their goods being depicted in machine-generated images without explicit approval. Otrium's terms with brand partners likely address these concerns, but the legal and reputational landscape remains unsettled.
Another limitation is workforce transition. Automated warehouses require fewer pickers but more technicians to maintain robots, software and grid infrastructure. Retraining legacy staff is possible but takes time and investment; some workers may face displacement, raising social and political scrutiny, particularly in regions with strong labour protections. A 2025 study by the European Trade Union Institute found that logistics automation in e-commerce reduced total warehouse employment by 12 per cent on average, though productivity per remaining worker rose by 35 per cent. Fashion companies must weigh efficiency gains against the broader employment impact and potential regulatory constraints on automation.
Finally, automation locks in certain process designs. Changing product flow, adding new categories or integrating a different order management system can be complex and costly once a robotic grid is installed. Flexibility, a traditional strength of manual operations, diminishes in a highly automated facility, posing a strategic risk if consumer behaviour or channel mix shifts unexpectedly.
| Dimension | Automated warehouse (Autostore) | Manual warehouse |
|---|---|---|
| Order cycle time | Under 30 minutes typical | 60 to 90 minutes typical |
| Inventory accuracy | Above 99.5% | 95 to 98% |
| Storage density | Up to 4x conventional racking | Baseline |
| Labour per 10,000 picks/day | Approx. 15 to 20 FTE | Approx. 40 to 60 FTE |
| Capital expenditure | EUR 5 to 10 million (mid-scale) | Minimal (racking and equipment) |
| Payback period | 3 to 5 years | Not applicable |
| Operational flexibility | Lower (fixed grid layout) | Higher (reconfigure aisles easily) |
Frequently asked questions
What should fashion companies do now?
- Audit current warehouse throughput and identify bottlenecks in receiving, putaway, picking and packing. Measure order cycle time, labour cost per unit shipped and inventory accuracy to establish a baseline for evaluating automation business cases.
- Model demand scenarios for the next three to five years, including peak seasonal volumes and potential channel shifts. Automation investments require sustained throughput to justify fixed costs; companies with volatile or declining order volumes should consider modular or leased solutions before committing to full grid installations.
- Test generative AI tools for product imagery and metadata creation on a subset of SKUs, measuring time savings, image quality and return rate impact. Establish clear policies on labelling synthetic media and secure brand partner consent where required by contract or regulation.
- Evaluate integration complexity with existing warehouse management, order management and enterprise resource planning systems. Robotic systems depend on real-time data feeds; legacy IT architectures may require middleware or upgrades that add cost and delay to deployment.
- Assess workforce implications early, including retraining programmes, potential redundancies and local labour law obligations. Engage employee representatives and plan transition support to minimise disruption and maintain operational continuity during commissioning.
- Compare vendor offerings on total cost of ownership, maintenance terms, software licensing and scalability. Autostore is one of several providers; alternatives include Exotec, Ocado Smart Platform and Vanderlande, each with different grid architectures, robot speeds and software ecosystems.
- Pilot before scaling by installing a small grid module or single picking zone to validate throughput assumptions, train staff and refine processes. Incremental rollout reduces risk and allows course correction based on real operational data rather than vendor projections.
Frequently asked questions
What is the typical payback period for an Autostore warehouse installation in fashion e-commerce?
Autostore reported in 2024 that typical customer payback periods range from three to five years, driven by labour savings, increased throughput and reduced real estate costs. Fashion retailers with high seasonal peaks and rising warehouse wages may see faster returns, while those with volatile order volumes face longer payback horizons.
How much can generative AI reduce product photography lead time?
Otrium reported that generative AI tools cut the median time to list a new SKU by approximately 40 per cent compared to manual studio photography. About You targeted a 50 per cent reduction in photography lead time for third-party marketplace sellers using similar AI-based image generation in early 2025.
Does warehouse automation reduce total employment in fashion logistics?
A 2025 study by the European Trade Union Institute found that logistics automation in e-commerce reduced total warehouse employment by 12 per cent on average, though productivity per remaining worker rose by 35 per cent. Automated facilities require fewer pickers but more technicians for robot maintenance and software management.
Sources
- FashionUnited: Otrium expands brand partnerships and infrastructure
- Autostore: Technology overview and system specifications
- Supply Chain Dive: Automated warehouses cut order cycle times in e-commerce (2025)
- Retail Dive: Generative AI in fashion product imagery (2025)
- ASOS: Annual Report and Strategic Update 2024
- Zalando: Annual Report 2024
- FashionUnited: About You marketplace automation initiative (2025)
- Autostore: Customer case studies and ROI analysis (2024)
- European Trade Union Institute: Employment impact of e-commerce logistics automation (2025)
