Posts Tagged ‘KU Leuven’

A high-level overview of the different PERSFO studies

Author: Robin De Croon, postdoc at KU Leuven

The PERSFO application is an ongoing, evolving smart food platform. Due to the global Covid-19 pandemic, the studies were set-up as hybrid studies.

Design study

The PERSFO application user interface was developed using a mixed-methods, user-centered design methodology and consisted of three focus groups, one co-design session, and one semi-structured interview with two dietitians. Ethical approval was granted by

the Ethics Committee of KU Leuven (G 2019 12 1911). The full design stage can is published in IEEExplore: Motivational design techniques to support healthy eating habits at work.


Data acquisition study

At the end of 2020, a virtual, but fully working prototype was deployed at Argenta Headquarters Belgium. The main goal was to test the technical integration and gather sufficient data to improve the recommender algorithm. The PERSFO application was connected to Sodexo’s food management system through the intermediate PERSFO architecture. The two-week study period resulted in 427 orders placed (participants could order multiple meals e.g., main course and snack), 9066 logged interactions with the app, and on average 9 orders per user. Seven participants registered their allergens, 14 participants liked meals with on average 7 ± 8 likes. Seven participants also gave feedback on the recommended recipe (by indicating disliked ingredients).


Integration Food Frequency Questionnaire and Quisper platform

In 2021, the integration of the Food Frequency Questionnaire (FFQ) and the technical integration with the Quisper platform received high priority. An improved prototype was deployed at Sodexo headquarters’ restaurant. Sixteen participants tested the application for one week and completed the full FFQ questionnaire. Participants received personalized feedback through the Food4Me Quisper service that was tailored to the Belgian food database. Moreover, the main aim was to gather sufficient data to tailor personalized recommendations based on the dietary profile of the user. User could indicate whether the agreed with a recommendation or not by using thumbs up and down icons.


Personalized, but transparent meal recommendations

Finally, at the start of 2022 the final version of the PERSFO application was deployed at J&J Beerse and J&J Geel in a fully hybrid study. The real menus served in the company restaurants are shown in the app, and users can order a meal when they are on location, or virtually confirm an order on the days they are working from home.


Participants’ FFQ data is processed by Quisper and the personalized feedback is used to recommend healthy meal alternatives for eight consecutive weeks. An important additional feature is the additional explanations users receive when they want to consult a personalized recommendation as shown below:

Right: home screen of the PERSFO application. User can click on ‘why’ to receive personalized advice why a certain meal was recommended. Left: the different weights are visualized towards the user.


The study is currently still ongoing and is expected to end mid-March. Over the two study locations, 81 participants completed the Food Frequency Questionnaire and 42 users are using the application to create (virtual) orders. In total 276 orders have been created, while 8600 interactions are logged.