Decoding cravings

x Sponsored content on Research Live and in Impact magazine is editorially independent.
Find out more about advertising and sponsorship.

Food retailers are under more pressure to compete than ever before – their rivals are no longer simply other retailers and grocers, but also new players and gig economy companies such as Uber Eats.

South African retail company Woolworths wanted to understand what is driving people’s new food preferences so that it could develop new products, change the format of its stores and promotions, and approach price from a new perspective.

“We needed to have a deeper understanding of how South Africans engage with food,” said Charlene Van Zyl, head of customer at the company.

Approaching the research using methods reported after the fact, such as traditional food diaries, wouldn’t have been effective, as engagement tends to fall off when customers no longer enjoy completing them, said Van Zyl, speaking at the IIeX Europe conference in Amsterdam. Additionally, the brand wanted to dig deeper into people’s motivations around eating.

Woolworths worked with Kantar to develop an approach aimed at understanding more about why consumers make the food decisions they do, and the reasons for those decisions across different categories and occasions.

Kantar designed a seven-day research study using an artificial intelligence chatbot diary, with the objective of collecting detailed information about what food people eat, and why. After recruiting participants through Facebook ad targeting, people shared their contributions through a chat conversation with an AI chatbot called ‘Serena’.

The chatbot was designed as female to reflect the brand’s customer base, 70% of which is female, and its scripts and questions were designed to reflect an outgoing ‘persona’, using first-person language.

‘Serena’ functioned as a coach, prompting participants to share photos and updates on their meals and snacks every three hours, then asking questions about whether there were any specific reasons why they had chosen that particular meal or snack aside from hunger, and the benefits over something else.

Karlien Kriegler, research director at Kantar, said: “We spoke to 1,100 people in total. The diary, focused on gathering habitual data in terms of when, what, how and where respondents were eating – and also provided meal preferences, as well as motivations of individuals over time. This enabled us look at rich behavioural patterns by hour, day, week and month.”

Gamification was used to incentivise participants, who were issued with a R50 (around £2.50 ) gift voucher daily, with an additional reward if they completed the full seven days.

“People enjoyed talking to Serena, and the high engagement achieved allowed us to collect a very large data set. This contained detailed responses, providing significant insights on meals consumed, combined with precise motivations for choosing those meals,” said Kriegler. 

New weekly waves of respondents submitted diary entries over one month. With more than 1,100 participants sharing information with the bot, there were more than 10,000 meal occasion conversations during the fieldwork, with three data points per occasion – when, what and why.

Kantar used artificial intelligence to process this large volume of data, and natural language processing to analyse the ingredients mentioned by participants into a list of more than 800; human researchers reduced this to 600. Additionally, the reasons people gave for eating particular meals were thematically coded.

Healthy eating emerged as a key motivation, with indulgence and cravings also featuring heavily. Throughout the process, researchers were able to pinpoint the ingredients related to indulgence – for example, chocolate, fruit and yoghurt – and map the motivation of indulgence and how it changed depending on the time of week or month.

According to the findings, participants were most likely to eat chocolate on a Thursday, when cravings for indulgent foods is climbing at the expense of healthier motivations. Chocolate and cinnamon also emerged as a popular choice for consumers.

The data also helped Woolworths understand trends from its transactional data. Previously, many of its promotions focused on treats and indulgence were carried out early in the week, but the research showed that people were in more of a healthy mindset at that point. In contrast, there was a lack of engagement with promotions for salad or healthy options pushed out on Fridays.

“The insights are broad and strategic, as well as practical and specific, and will be used in multiple areas of the business,” said Kriegler. “With the data collected by Serena, we will be able to inform marketing, pricing, product development and trade teams. There are significant applications from this one rich data set.”

The impact of time of week and month on food choices

Monday, mid-month
Cravings are very low and healthy eating is very important.

Thursday, mid-month
Craving climbs steadily throughout the week, and indulgence increases, with healthy eating becoming less important.

Saturday, end of month/after pay-day
Craving at its highest level.

This article was first published in the April 2020 issue of Impact.