Too Good To Go, a marketplace that connects consumers with businesses to combat food waste, has leveraged advanced artificial intelligence to analyze feedback from its partner stores. The initiative, executed with Amazon Bedrock and AWS Partner Mistral AI, sought to understand why certain stores ceased using the platform. By examining thousands of interactions monthly, the company gained valuable insights into issues affecting partner retention, which in turn facilitated targeted product improvements.
Operating a platform where bakeries, grocery stores, and other food businesses can sell surplus food at discounted prices, Too Good To Go serves over 100 million registered users and nearly 200,000 active stores. The company has successfully averted over 400 million meals from being wasted. However, when some partner stores stopped supplying or exited the platform, Too Good To Go faced challenges in identifying the underlying reasons.
The existing processes for conducting root cause analyses were inefficient and lacked depth, prompting Too Good To Go to explore its sales team’s interactions with stores that had temporarily ceased their supplies. These conversations were seen as crucial data points that could inform improvements in both product and marketing strategies.
To address these retention issues, Too Good To Go developed a solution employing Mistral AI’s models for analyzing and categorizing the reasons behind stores discontinuing their supplies. Built on Amazon Bedrock, a fully managed service offering various high-performing foundation models, the solution utilized data extracted from sales team interactions. The team categorized common issues into broad categories and manually processed a subset of entries to create a training set for the AI. Through rigorous prompt engineering and iterative refinement, they achieved an impressive accuracy of 85–90 percent in categorization.
The selection of Mistral AI within Amazon Bedrock was primarily due to its multilingual support, strong categorization performance, and cost-effectiveness. “The main advantage of using Mistral AI models within Amazon Bedrock is that we had a lot of confidence in the system we were working with,” said Daniel Redgate, product analytics lead at Too Good To Go. “We didn’t have to worry about data being sent to an unknown service provider to an unknown location. And we knew that if we wanted to ‘productionalize’ the model long term, then it would be much easier working within the AWS environment.”
The implementation of this AI-driven solution generated significant insights, prompting Too Good To Go to reevaluate its traditional onboarding processes. Redgate noted, “We’ve identified a completely new category of stores, which will require a different onboarding process and potentially different touch points with our sales team.” This new category was found to represent about 20 percent of the analyzed tickets, necessitating a fundamental overhaul of the company’s approach to onboarding.
Furthermore, the analysis identified that many stores experienced difficulties with the platform’s scheduling feature, especially during vacation periods, which led to unintended inactivity. In response, Too Good To Go plans to enhance communication regarding product features and consider modifications to its user interface to improve intuitiveness.
Looking ahead, Too Good To Go is exploring innovative applications of AI, including image analysis to enhance consumer transparency and optimization recommendations for partner stores. This ongoing evolution underscores the company’s commitment to leveraging technology-driven solutions in its mission to reduce food waste.
See also
Germany”s National Team Prepares for World Cup Qualifiers with Disco Atmosphere
95% of AI Projects Fail in Companies According to MIT
AI in Food & Beverages Market to Surge from $11.08B to $263.80B by 2032
Satya Nadella Supports OpenAI’s $100B Revenue Goal, Highlights AI Funding Needs
Wall Street Recovers from Early Loss as Nvidia Surges 1.8% Amid Market Volatility
