GO INSIGHT
AI & Automation
AI-assisted tools that help organisations monitor complex environments at scale. Leveraging automation for better efficiacy and surfacing what matters without the manual overhead.
CASE STUDY
Beyond the Marketplace: Automating Intelligence on the Online Donkey Skin Trade
Go Insight collaborated with The Donkey Sanctuary to support their on-going efforts to generate intelligence-led insights of the global donkey skin trade.
Centralising Intelligence through Technology
The global trade in donkey-derived products from raw skins to processed ejiao is transnational and complex. Traditional monitoring methods struggle to track activity across multiple platforms, languages, and regions, leaving critical gaps in understanding trade patterns.
Go Insight addressed this challenge by developing a technical approach to centralise and automate intelligence collection we reduced manual effort while improving the consistency and structure of collected data.
Actionable Intelligence
We tested whether automated monitoring could reliably track online trade activity at scale. This included identifying which platforms could be scraped, how to detect relevant listings across multiple languages, and whether automated tools could consistently produce usable intelligence.
We then tested whether automated monitoring could reliably track this activity at scale, using Python-based scraping frameworks alongside browser automation to assess scalability, relevance, and repeatability. Across the initial test platforms, well over a thousand listings were scraped with a relevance rate above 60%, confirming that automation works when guided by human-led platform selection and keyword strategy.
Iterative Testing and Insight Development
Manual analysis identified seller behaviour, refined multilingual keyword strategies, and interpreted regional terminology. Automated tools were then deployed to extract structured data while navigating platform-specific barriers, with testing revealing significant variability between platforms that directly informed platform prioritisation and tailored configurations.
To improve accuracy and reduce manual review, we introduced a two-stage classification system: a keyword classifier handles clear-cut cases, while an AI model (Llama 3.3 70B, via Groq) assesses borderline listings in context, recording a written explanation against every decision for full analytical transparency.
Building a Sustainable Monitoring Solution
This feasibility work established a blueprint for scaling automated intelligence collection across global online markets. Which resulted in a fully operational, web-based scraping and classification application demonstrating that AI-assisted monitoring of complex online trade is achievable at low cost. The system monitors platforms across three language environments, automatically fingerprints every listing to flag new market activity. The results are presented through a dashboard giving the TDS team at-a-glance view of new and returning activity alongside exportable intelligence.
The result is a monitoring framework designed to adapt to evolving digital environments and a blueprint any organisation can build on to turn fragmented online activity into structured, actionable intelligence.