FSA
HELPING THE FOOD STANDARDS AGENCY TO KEEP THE PUBLIC SAFE
CLIENT
REQUIREMENT
The Food Standards Agency wanted to understand how they can prioritise the inspection of food establishments without compromising public safety.
CHALLENGE
Working with sparse and messy data, a lot of which was from handwritten notes, to extract useful information out of them and create a predictive model presented a true challenge.
SOLUTION
Our risk segmentation model forecasts how well the restaurants are likely to follow food safety standards, allowing scarce public resources to be allocated more efficiently.
Optimise, not compromise!
You eat out because you want convenience, variety or a gastronomical experience. You assume it is safe. However, your assumption is based on the restaurant having been checked it meets certain quality standards.
The Food Standards Agency (FSA) is the guardian of the health and safety of food establishments, and they introduced the familiar 0 to 5 Food Hygiene Rating. The Local Authorities visit every new site to check the quality, which takes a lot of effort. But can the FSA and Local Authorities optimise the inspection strategy to get better, faster results?
No food for thought – working with sparse data
Can this effort be better targeted to focus first on those that we can show are more likely to get a low rating? If so, how do you prioritise? Is there a set of the indicators that can help to foresee the food establishments that are less likely to follow good hygiene standards? Are these indicators even known or recorded anywhere? Facing these questions, Decision Lab had to work with very few data available on food establishments. Going through the unstructured sources such as the scans of handwritten notes that varied between the Local Authorities was not an easy task.
The Risk Engine
Having collected details on a large sample of food establishments we developed a risk engine. We developed a set of the risk indicators and used data science methods to rank the establishments against the FSA’s food hygiene requirements. The results can enable Local Authorities to identify the highest priority checks and optimise their inspection regime.
Clever data science that provides explainable results
We developed the Risk Engine using a variety of machine learning algorithms, and cutting-edge model interpretation libraries. We focused on models that were easily interpretable and so the forecasts of which establishments are likely to be compliant and which can not be understood and trusted.
A win-win for all
As part of the ‘Regulating our future’ programme, Decision Lab’s Risk Engine assisted FSA achieving its goal of developing a modern and resilient approach to risk food establishment regulation, by allowing resources to be planned efficiently whilst helping FSA to keep the public safe.
Let’s collaborate
Send us an email, to discuss a new project.
We’re a team of innovators who are excited about unique ideas and help companies to create amazing solutions.