TASTE PROFILEs 🌽
Gousto is a recipe box company on a mission to become the UK's favourite way to eat dinner. Customers choose from a range of recipes every week and are sent the exact ingredients to cook these recipes direct to their door. This helps to reduce food waste whilst bringing the joy and excitement back into cooking.
What do you we currently know
When looking at the data we saw that 89% of customers are already satisfied with our variety. However, the lack of choice is the number 2 reason to skip a box currently. We hypothesised that this might be down to the fact that customers were struggling to find new recipes or inspiration within the menu. From user testing it was highlighted that this might be down to poor filtering experience or the lack of engagement within the app, and therefore simply satisfying only a sub-set of our customers isn’t enough.
Therefore with the PM we decided to create two OKR's to measure how we might measure customer satisfaction score.
KR1. Unlock variety by +40ppt for customers who are extremely satisfied with variety (from 34% to 74%)
Drive retention with +13ppt MCVR (60 to 73%)
I recently joined the Turnips team where I worked with a Lead designer, PM and Junior Engineer. My role was to get involved in the discovery process around what a taste profile experience could look like within Gousto. I was tasked with digging deeper into the exploration of how the experience could evolve and entice users to reuse the experience after 3 months. I created various Low Fi mock ups, High fidelity designs, Figma prototypes, as well as working with the engineers to build higher fidelity prototypes.
wHO ARE GOUSTO?
It was decided by the business that the first half of the quarter would be spent in discovery, to find out what concepts worked best with customers as well as get buy in from senior stakeholders.
I began researching into full screen, quick fire interactions to gather inspiration and understand best practice techniques for increasing customer retention. With Gousto being a subscription service I looked at various online subscription companies, as well as popular apps, that have huge engagement rates. This library of inspiration was then used to inform and inspire ideas during an ideation workshop.
I ran an ideation workshop with our cross-functional team focused on how might we best create an engaging and interactive way to showcase choice and variety. We focused our attention on the mobile experience, as 80% of our customer base were App users. Therefore how could we utilise touch and interactivity patterns to prevent choice fatigue and customers dropping out the of experience.
Workshop structure (45 mins):
Overview of previous learnings from the menu team (radishes)
Competitor inspiration - what makes a good experience? What examples are that showcase this?
Crazy 4's (Pairing solutions) + affinity mapping
Crazy 4's (Single solutions) + affinity mapping
Create fun and engaging experience
From the cross functional workshop, and the ideas that we voted for as a team I decided to take these and start fleshing them out. I set aside some time to do some quick scamp drawing with pen and paper. I later stuck these on the wall and ran through them with the team and refined them down with the lead designer.
PAIRING WITH BRAND
We made a refined list of ideas from the workshop as well as my quick scamps to further develop which were focused on choice and variety. We ensured these ideas were as varied as possible to maximise learnings from the user testing. These ideas were then combined with different ideas of how to showcase different ways a customer might interact or choose which recipes they found to be most appealing. It was important to consider brand, to make sure that the look and feel as well as the language we were using felt consistent.
Idea 1 - Similar to Tik Tok, infinite scrolling paradigm where users either scrolled up for recipes they liked or down for ones they didn't.
Idea 2 - A Tinder style left and right mechanism, where the recipes where displayed on "stacked" cards
Idea 3 - A "this or that" approach where we show cased two recipes at once for users to choose between.
We used Figma to explore these ideas, with brand bringing in potential image assets and writing variations of copy and myself focusing overall layout and key interactions. We quickly discovered that Idea 1 was very overwhelming, idea 2 felt very "off brand" and weren't keen on following a tinder-esq interaction. Therefore we decided to take idea 3 of "this or that" into further exploration.
MODERATED TESTING with a prototype
In order to understand how customers would interact and gain from this experience I decided to conduct some moderated research. I wanted to get a better understanding of user expectations, did they understand the information they were being shown and what did they think would happen when they finished the experience. We ran this study with 10 participants.
I worked closely with our frontend developer to produce a functional prototype to show to participants. For this test we decided to focus on the recipe battle, where we were pairing two recipes against one another.
We sent the link to participants and asked them to share their screens with us and we would simply observe what they did until the end of the test, where we then dug a little deeper into questions.
Key themes from the research
The first screen showed two CTA's "let's play" and "Dietary requirements"- many participants were confused by the language as for users who were vegan did not consider themselves to have dietary requirements
Users started to disengage after about 5/6 recipes cards, there needed to be better feedback on how long this would take
Almost all participants thought that after this experience they would then see a tailored menu based on the recipes they had picked.
Most participants commented saying that they found the imagery to be very colourful and visually rich
From the research we found some really interesting insights and some take aways to explore further with regards to the design. The core questions we decided to explore were:
Questions to Validate:
Is the distance from title and centre of the image too far away?
Because its full screen, could user think that they can go back to previous recipe?
Does the full version feel like they are forced to have to pick something? - What if they don't like either?
Does visibly being able to view the stack make the experience feel lighter?
Should there be some sort of counter to show how many recipes they have left (endowment effect)
Should there be more feedback when a user picks a recipe they like?
Therefore I decided to iterate on the experience a little further. We decided to make the font size larger so it was much clearer what recipe they had chosen. We included a "skip this recipe" button to allow users to not have to choose anything if they didn't like either recipe. From the testing we noticed users weren't 100% sure whether they have liked or disliked a recipe, therefore we wanted to included some visual feedback so users felt it was more playful. Gousto uses very playful language such as "yum" "Give it some", we decided to try to inject the brand further into the project. Lastly since we only decided to test with 10 applicants we felt that wasn't a big enough sample size to fully validate our hypothesis. Therefore we decided to run the updated designed in a survey of 200 people.
EXPERIMENT Vs baseline
From the 200 people that complete the survey, we found we had very high engagement rate for the first 5/6 recipes then customers tended to start dropping out. We noticed that there seemed to be an optimal customer to send the experience too, of users who has order box number 2 or 3, any earlier than this we have very little engagement, likewise for customers who were on box order 5+. Therefor from the baseline we can confidently say that there was an uplift between battle and baseline up until box 2 - box 3 is significant if we decrease our confidence level to 90% (from 95%).
Everything nicely aligned with the bayesian approach above that performed and it looks like the recipe battle is definitely a winner here and had great engagement with users. The survey data seems to have no impact on basket match above box 4 for all the variants.Chart below - the lower the p-value the more confident we are that the difference between basket match is “real” and not just a random effect.
From our early learnings from both the Moderated testing and the survey we felt confident to roll this out to an A/B test.