Introduction
The problem of food wastage has many dimensions and layers. Although a lot of food is wasted on each level of the food supply chain, the consumers can help to reduce it by changing their habits and attitude to the food.
Many consumers use shopping lists, recipe apps and recommendation services, but they still buy more than they can eat. The reason behind it, as the research results show, can be an emotional hunger. It was also obvious that the food turns into “food waste” in the consumer’s fridge. So the problem my solution was supposed to solve was how it could make products the users already have in their fridge more attractive and, in a way, add to their emotional value for consumers.
The solution I found was FoodGuru - an app with a chat bot that will give tips, recipes and recommendations on how to use products in the fridge to cook something interesting while reducing the food waste at a household’s level.
The usability testing showed that the concept was received positively by users and had a potential to grow. I used the Design-Thinking Process which allowed me to keep focused on the users’ needs.
Secondary research
The problem of food waste is high on the agenda. According to the Food and Agriculture Organisation of the United Nations, “an estimated 1/3 of all food produced globally is lost or goes to waste.” FAO’s report “The state of food and agriculture 2019” shows that the problem of food waste at the consumer level is typical for high-income countries as “the higher the household wealth, the more consumers will waste.”
Norway as a high-income country is not an exception. According to the official statistics, Norwegian households waste around 43 kg of food every year, which means that each person throws away about 800 g of food a week.
To better understand the general patterns and motivations behind consumer behaviour as well as reasons that lead to food waste in Norway, I analyzed a TV show called “Matsjokket” (“Food shock”) on NRK. Although people don’t usually behave naturally on TV, the show does highlight typical behaviours of consumers as its purpose is to raise awareness and urge people to change their habits.
The analysis of the consumer behaviour in the TV show gave the following insights (see file “TV show - Data capture sheet” for detailed findings):
Primary research
Although the secondary research helped to understand some of the reasons why consumers waste food and outline the target audience, it was still unclear how people make decisions and what drives their motivation. Besides, some questions remained unanswered.
To find out answers to these questions and collect more data about consumer behavior I conducted a number of in-depth user interviews. As I needed to collect qualitative data, I chose a descriptive approach and conducted a set of user in-depth interviews. The interviews were semi-structured to cover the questions I already had and to give the participants a chance to tell more about their experience. The interviews were conducted remotely (phone interviews) due to social distancing.
To select the participants the following criteria were outlined:
Key insights
Having analyzed the data gathered during the research, I synthesized it into the following insights:
1. Grocery lists don’t always help to avoid impulse buying. Most people admit that they will buy more than there is in their list, but they don’t like to believe that they buy impulsively.
2. Spontaneous purchases don’t always generate food waste. All kinds of participants - those who claim that they never throw anything away, those who think they throw away just a little and those who feel like they throw away a lot - have similar shopping patterns. The difference is that those who don’t waste food consciously decide not to throw it away and have established routines aimed at using leftovers.
3. There is a clear connection between the amount of food waste and the negative feelings of the participants: the more they think they waste the more negative emotions they experience. It’s important to mention that the participants don’t have information about how much food they really waste, it’s their own judgment.
4. People don’t always buy food because they need it. Sometimes it’s an entertainment or a treat. They are interested in discovering new products and tastes and in new impressions. And they don’t see any problem with it.
5. Despite the unstable economic situation people are not planning to buy cheaper products to save more money, at least at this stage. The main reason why they change their shopping habits is the priority of health, not the financial insecurity.
6. Even though many people know that certain products can be as good after the “best before” date as they were before it, they still associate a product that stays in the fridge longer than acceptable in their opinion with a health risk. Sometimes they can give such products a chance and smell them, but emotionally they are not ready to use them. This mostly applies to open dairy products (milk, sour cream, cottage cheese, etc) or wilting fruit and vegetables.
It seems that the “best before” date plays the role of an emotional “threshold” for consumers. If the “best before” date is approaching and consumers notice it before it comes, they are willing to save the food and extend its shelf life.
7. The majority of consumers see their health as the highest priority. In terms of food “healthy” means “fresh”, but “fresh” is directly linked to “best before” or “use by” date. Thus, the closer it gets to the expiry date, the more “dangerous” and “unhealthy” the product seems to the consumer. If the product was open, then it is considered “dated” and “unsafe to eat” even before the “best before” date. For example, for consumers dairy products become “not safe to eat” within 3 days after opening a pack (see №6 for the quote).
8. Meal planning doesn’t always work because people tend to change their preferences. The more “stakeholders” in a family (for example, kids whose opinion is taken into account), the higher the chance that the plans will change.
Personas
Based on my research, I created two personas - Emily and Maria. Emily belongs to those who don’t think they waste a lot of food, while Emily represents those who feel guilty when they throw food away, especially if it’s unopened.
To better understand the user’s pain points and how food turns into food waste I summarized the most common patterns into the following scenarios. The food is marked with three different categories: “Fresh” - the best quality, “Used” - acceptable and safe, but not the highest standard, and “Bad” - inedible, unsafe, to be thrown away. They don’t represent the real condition of the product, but how the consumers feel about it:
The scenarios above show that consumers can start considering food “bad” or “of low quality” long before the expiry date. In fact, in the eyes of consumers the fridge is the place where their food goes from “fresh” to “bad” within days, especially if it came in a pack and the pack was opened.
To better understand why the food is being left and forgotten in the fridge I combined insights 4 and 8 and used 5 Whys Technique:
It’s worth mentioning that all the participants, despite their different habits and attitude towards food waste, used the word “interesting” or “unusual” at least once during the interviews in relation to food.
So it can be concluded that the food doesn’t only satisfy physical hunger, but also gives consumers emotions and impressions - they can be curious when they buy a new product, a new dish can entertain them with its unusual or interesting taste, etc. From this perspective it seems logical that as soon as the product is in the fridge it starts losing its novelty and the emotions the consumer was expecting fade away. It also works in a family where all members have their own expectations and preferences related to food. Based on this the problem statement can be defined as:
Ideation
Having conducted an ideation workshop , the participants have decided that the final solution should include the following aspects to satisfy the users’ needs and to address the food waste problem at the same time (Image 1).
In scenarios from the previous chapter the product will be used at the “Fridge” stage and work as an “advisor” in the same situations (Image 2).
As the solution is supposed to deal not only with physical, but also with emotional hunger, while giving our users ideas what to do with the food they already have in the fridge, it could address the users need in the key path scenarios in the following way (Image 3).
Based on the scheme above the solution should meet the following requirements:
Having summarized the data and ideas discussed during the ideation stage, it was decided that the best solution, in this case, will be FoodGuru - an app with an AI-powered chatbot that will give advice and guidance to the user every time he opens the fridge and decides what to cook with the food he has there or what to do with certain items.
The user will be able to choose a chatbot personality - an influencer from the food industry - that will establish an emotional connection with the user and act as a role model and a credible source of information. The app will also be able to share some of the results (e.g. prepared meals) on social media to create awareness and competition.
Information architecture and user flow for the onboarding process
The chatbot will function according to the following general algorithm:
After onboarding the chatbot sends a greeting and the user enters the name of the product or its photo, the AI recognizes it and identifies the category - “Leftovers”(a meal that was cooked at home and not eaten), “Vaccum or antimicrobial packaging”(perishable products that are sold in a special packaging - like meat, fish, dairy, canned food, etc, and that has to be consumed within a short time after the opening of the package) and “Can be sold loose”(products like vegetables, fruit, nuts, etc that can be sold without packaging). After identifying the category the chatbot gives recommendations on what to do next.
Low-Fidelity wireframes
For the chatbot I used the prototyping tool Botsociety and wrote scripts according to the algorithm of the chatbot flow.
Prototype
For the expert review of the prototype I used Jakob Nielsen’s 10 Usability Heuristics. The prototype is available at the link https://xd.adobe.com/view/8fa7f9f2-f04f-4060-4270-8096f6a88f13-f996/
Mid-fidelity wireframes
Test
To test both the signing in process and the chatbot the usability testing was conducted in two stages. For the first stage the digital prototype created in Adobe XD was used. For the second one two chatbot scripts were developed using Botsociety. The testing sessions were conducted via Whatsapp, as it is a quick and cost-effective way to test the chatbot before the high-fidelity prototype could be designed. As the chatbot will have an interface which is typical for messengers, testing it in Whatsapp will give the user more freedom to act naturally. To simulate the bot we used a human participant that followed the script manually.
Goals:
The following goals were set up as primary for the testing:
- Ensure that the signup process is easy and quick
- Gather feedback about the idea of a chatbot giving them advice
- Identify any roadblocks or design issues
Test objectives:
1. Onboarding and Sign-Up:
The objective of the test will be met if the user manages to proceed from the splash screen to the chat page creating an account via email or Facebook. It will be tracked how many users will try to skip onboarding.
2. First interaction with the chatbot:
The test objective will be met if the user manages to proceed from the welcome message in the chatbot to a recipe with the product he mentioned. It will be observed if the chatbot’s logic is clear to the users and if they understand in general how to interact with it.
3. Getting a tip from the chatbot:
The test objective will be met if the user manages to get a tip from the bot and proceed to the recipe. During the session the feedback will be gathered to understand if the script was clear and if there were some major roadblocks in the chatbot flow.
Method:
Usability testing will be conducted remotely using a digital prototype and messengers (WhatsApp and Facebook Messenger). The sessions will be moderated via a voice call.
The results of the test were summarized in the following tab.
Key findings:
#1. The users in general liked the idea of having a chatbot as an advisor and an entertainment in the kitchen. They also found that it could be useful in the countries where there is a confusion over “best before” and “use by” dates.
#2. The users who read the instructions were less annoyed when the chatbot couldn’t meet their expectations. But the problem is that many users never read any instructions, so for the situation like these (see the picture below) the bot should be programmed to urge the user to clarify what he means:
Design opportunities:
- How Might We make the chatbot ask the users to clarify the request?
- How Might We enable the chatbot to recognize the natural language of the user rather
than giving them a number of predefined options?
- How Might We modify the script so that even the users who don’t read instructions could easily recover from errors?
This example shows that though AI wasn’t a part of this testing (but is mentioned in the requirements), it can play a crucial role in the user experience of the future app. It’s also important to test and refine all the scripts of the chatbot before they could be integrated into the app.
Another example of the fact that users tend to skip instructions is that all the participants tried to skip the onboarding. As there was a short introduction those who actually skipped felt a bit confused further. One of the solutions to this problem was to give the user a chance to come back and reread the introduction by pressing the “back” button. Another way to address this problem would be to add a slogan or a short logline to the first screen, so that the user could have some idea about what is expecting him ahead.
#2. When users don’t understand the final goal of the task, they can get annoyed. During the test the users didn’t get why they needed to take all those steps that the tip “How to check if the lamb is fresh” included. They would prefer to have an overview at the beginning and then decide if they need it or not.
#3. The users wanted more details about the chefs' personalities and what kind of help they could offer. Although for the low-fidelity prototype a detailed description of each chef wasn’t planned it could have been one of the future iterations.
#4. It seems that the users see the app as a source of verified recipes. They feel that the app is a quicker alternative to googling recipes. A recipe that will allow users to use or save the ingredients they have in the fridge is seen as an ultimate goal. So they want to get there as quickly as possible. After getting the recipe they lose interest till the next time they need to interact with it.
Conclusions and recommendaions
The usability testing proved that the concept of a chatbot that can help in the kitchen is viable and made a positive impression on users. Despite the fact that the prototype wasn’t perfect and had limited functionality the users understood the general idea and the value of the app.
Testing the chatbot via messengers with a human operator also proved to be quick and efficient. As the operator is supposed to follow the script, interaction with users immediately highlights pain points and roadblocks in the flow.
Based on the feedback from the users the main values they see in the app are:
Practicality - A quick way to get a recipe and use what you have in the fridge reducing the food waste
Exclusivity - A way to feel special when you get a recipe from an expert or an influencer
Human touch - A humanized digital experience that makes you feel like you are chatting with a real person
The further steps in research should help to identify what is the primary value for the target audience and in what proportion they should be combined in the solution. For example:
