Jailson Soares

Vivian Ribeiro

Harry Biddle

To enable product traceability in the soy supply chain, we must move beyond the traditional focus on producers and consumers to address the ‘hidden middle’ – the vast and complex network of silos, warehouses and processing facilities where visibility is often lost. Without detailed data on these facilities, companies buying soy will remain blind to its origins, as mixing of different batches masks potential environmental risks. This information gap also limits the ability of governments to enforce farm-level measures to tackle deforestation such as the EU Deforestation Regulation (EUDR).

Understanding the precise location, ownership and capacity of these facilities is the key to product traceability. Knowing whether a facility in Brazil is an on-farm soy silo or a high-capacity regional terminal determines the feasibility of different chain of custody models, from Identity

Preserved (IP), which tracks soy from a single origin, to segregated, which keeps certified soy separate from conventional soy, to mass balance, which allows mixing while accounting for certified volumes. With detailed information on these facilities, stakeholders can identify where supply chain mixing might occur and distinguish between transparent direct-sourcing routes and higher risk indirect paths.

Current transparency measures in the soy sector are limited. The most advanced is Brazil’s registration system for agricultural storage units, known as SICARM (

Sistema de Cadastro Nacional de Unidades Armazenadoras

). However, manual registries are slow and cumbersome, and struggle to cope with Brazil’s vast size. They contain significant gaps, particularly upstream where thousands of private on-farm silos and temporary silo bags are not registered. Moreover, in the global market for commodities, creating separate national databases is an inherently fragmented approach to increasing transparency.

What is needed is a proactive, international system that supports collective efforts on product traceability and harmonisation across countries. The 

 for palm oil facilities shows what is possible with better information. The list provides a global map with standardised identification for over 2,300 facilities, enabling companies to 

align deforestation-free protocols and sustainability reporting

Trase has developed a solution that uses the two state-of-the-art geospatial foundation models – 

 – to automatically identify agricultural facilities at scale. The technology enables Trase to scan vast agricultural areas to detect the distinct physical signatures of silos and warehouses that were previously unknown. 

 helps identify facilities, including those in remote areas, providing companies and governments with the most accurate picture to date of where soy is actually being produced, stored and processed before it reaches the global market.

Trase can not only identify the location of a facility, but also its specific function by using a sophisticated AI workflow that blends wide-area discovery with high-precision analysis. While AlphaEarth conducts broad prospecting to flag potential sites across vast landscapes, Clay performs detailed structural evaluations to distinguish legitimate soy facilities from other rural buildings. By automatically recognising key visual markers such as cylindrical vertical silos, facilities layout, flat warehouses with arched roofs, and even temporary silo bags, this system transforms high-resolution satellite imagery into a validated map of the soy supply chain’s logistical links.

Trase also provides important information on the ownership of facilities by integrating a large language model, 

 to connect physical assets with corporate entities. Currently, we can identify the ownership of 40–60% of facilities and are working to improve this rate. This system scours unstructured data from corporate registries, tax databases and digital maps to find ownership links that are often missing from official government lists. Advanced fuzzy-matching algorithms then synchronise these noisy datasets, linking facilities even when company names are misspelled or inconsistently recorded across different sources. This synthesis transforms messy, fragmented web data into a structured, usable database that identifies not just where the soy facility is, but who owns it.

Through these innovations, Trase has identified hundreds of soy facilities in Brazil that were previously missing from existing registries. With these additional detections on top of official databases such as SICARM, Trase has expanded and validated the known universe of Brazilian soy facilities to over 9,300. Trase’s new soy facilities map 

 with mapping slaughterhouses in Brazil, cocoa cooperatives in Côte d'Ivoire and palm oil mills in Indonesia.

Trase is working to address remaining challenges. For example, confirming whether a facility is operational still requires a combination of AI inference and periodic inspection. Furthermore, unraveling complex corporate ownership structures remains a difficult task that requires human knowledge along with automated reasoning. Recognising these limitations is crucial as we move from simple detection to more operational intelligence.

Despite these obstacles, the vision of a global database for all commodities linked to deforestation is closer than ever. Trase’s use of spatial data and AI offers a scalable solution that can be applied beyond the Brazilian soy sector, encompassing other high-risk commodities such as beef, cocoa and palm oil worldwide. The creation of a transparent, freely accessible data-driven facilities map is a crucial piece of the puzzle to finally decoupling global commodity trade from deforestation.

Explore Trase’s Brazil soy facilities map