"allplants is a plant-based meal delivery service operating with a subscription model. As part of the allplants team I had the opportunity of working across retention and acquisition channels alike. Currently my main focus is the Conversion Rate Optimisation of the e-commerce shop experience + supporting the operation and marketing teams with the launching of new dishes and campaigns"
  • UX design
  • UI design
  • Research
  • Usability Testing
  • Figma
  • Amplitude
  • Optimize
  • Miro
  • High fidelity prototypes
  • Strategy
  • A/B tests
  • User flows
  • Ongoing

Onboarding Wizard 1.0

Problem statement

The goal of this project was to increase the number of customers going from box 1 to box 2.

Users & Audience

Initially, the user of the wizard would be any customer making their first order in the allplants website.

Roles and responsabilities

3 people: Retention Product Manager, Product Designer (myself) + Marketing Manager with the occassional collaboration of a Front End Engineer

The constrains

One of the principal constrains of this project was that it had to be developed with minimum involvement (or not involvement at all) of the engineering team.

Another factor was that although the project was a retention initiative it took place on the acquisition part of the website therefore it was of vital importance not to damage conversion rate at any point.

Get + users to "eat and repeat"

When this project started the filtering capabilities present at the website weren't enought for users to find the most appropiate meals to their taste, which meant the selection of meals could be a "hit and miss".

Starting with the idea of having a tailored allplants experience will make users more likely to repeat their purchase, the team set up to create an "onboarding wizard" to facilitate the users initial decisions and gather insights about their motivations for the marketing team

The process

First we ran a workshop on miro to ideate and align the outcomes that each part of the team wanted from the onboarding wizard.

Following Lean Agile methodologies we came up with a series of experiments that would allow us to test our hypothesis quickly and with minimal effort.

Our main goal at this point was to confirm or discard the hypothesis that people prefered to be told what to choose rather than choosing themselves.

The riskiest assumption

Do users even want to take a quiz? Would they trust our recommendations enough to continue with their purchase? What would happen if we give everyone a box with the best sellers vs if we tailored results to users needs? (even if they are not the most popular dishes) Would any of these experiments damage CVR?
Hypothesis 1
  • We believe enough people will want to take the quiz
  • We believe having a quiz (even if imperfect) will not hurt CVR
  • Initially we though of making a smoke test but decided againts it as it could hurt CVR
  • Instead we created a "minimal effort" quiz
  • Embedded typeform without tailored questions + email capture. Take users to shop with a pre-populated basket with the most popular meals.
  • Everyone gets the same result
  • CVR of shop not decreasing
  • % of people clicking "Go to quiz"
  • % of people completing the quiz
  • % of people who complete the quiz reaching checkout
Hypothesis 2
  • We believe we can get more people to box 2 by figuring out their spice level preference.
  • We believe those receiving meals that are best cooked in the available appliances will be more likely to reach box 2.
  • Create a new test with single choice type of questions for spice level + cooking method
  • A/B test untailored box recommendation vs tailored quiz
  • Number of people reaching second box should be higher for those with tailored result vs untailored result
  • % of people completing the quiz
  • % of people who complete the quiz reaching checkout


The experiments didn't damaged the general CVR of the website.

Although the Click Through Rate to the onboarding wizard was lower than other call to actions on the website, its covertion rate was higher than other routes.

After running the test for a few weeks and observing the behaviour of the cohorts the data proved that those who had the personalised box were more likely to reach box 2 that customers that have the "most popular dishes" box and than those who didn't do the onboarding wizard

At this point we added questions for segmentation targeting of the users for email campaigns and allowed customers to change the selection of meals recommended by adding a "customise" button.

The use of engineering time was only necessary for the creation of the results template, the url containing different meals could be created by anyone of the team, and the onboarding wizard ran completely on typeform's logic.

The version 1.0 of the onboading wizard had a total of 24 unique results and covered questions related to cooking method, spice, allergens and motivations. It collected over 15k responses and was successfully used on retargeting leads and abandoned basket flows. The project remained in the hands of the marketing team, as I focused in different priorities. The version 2.0 of the onboarding wizard would resume a few month later as acquision team initiative.

Onboarding Wizard 2.0

Problem statement

The goal of this project was to use the onboarding wizard to improve CVR, increase the number of people completing the flow and to move away from typeform.

Users & Audience

Leads, prospective, abandoned basket, churned and new customers alike.

Roles and responsabilities

3 people: Acquisition Product Manager, Product Designer (myself) + Front End Engineer.

The constrains

The new onboarding wizard was to be launched as part of the January Campaign. The campaign had new branding guidelines and messaging based on the premise that is ok to crave non vegan food and to "cheat" on veganism.

Given the number of changes that were to be made on the wizard; we first needed to learn if we were making the right decisions via a typeform quiz before creating our own onboarding wizard from scratch.

Generating Delight

Althought we had proven that those who finish the quiz were more likely to convert, the number of users that were dropping off was high (over 50% in some cases)

The fact that the quiz was set on typeform meant that questions couldn't be changed or reordered without breaking the logic, long QA processes and the need for manual updating if one of the recomended dishes went out of stock.

We needed an onboarding wizard that felt part of the allplants experience.

North star workshop

Starting with data we had collected from the first onboarding wizard, we ran a workshop with the different members of the team to define the primary and secondary outcomes we would like to achieve with the new onboarding wizard.

Once we had defined primary and secondary desired outcomes we set up a matrix that would help us set an number of experiments and the key results we will measure on each.

After that we created wireframe for the new onboarding wizard that could be created on typeform and set the UI we could use for each question for when start building our own.


We set up templates for each one of elements of the wizard:

  • Progress bar
  • Navigation
  • Question content
  • Results page
  • Customisation flow

And all the different question types:

  • Single choice with "animated" illustration
  • Multiple choice text
  • Text input
  • Single choice with image
  • Display of text input
  • Result page

To make traking of dropoff points easier to identify we took the decision of displaying one question per page for the final designs.

Final Designs

During an internal alpha release of the onboarding wizard it came to our attention the some users had problems differenciating between the navigation, progress and content of the quiz given the reduced colour palete of white, black, green an orange

The styles were updated to make the content of the quiz obviously apart from the rest of navigation, error and information messages and retested.

The current version of the onboarding wizard constains 12 questions that capture diet, intent, motivation, spice level, frequency, portion size and email.

The Outcome

With aprox 3000 new customers going through it monthly, those who start the new onboarding wizard today are 75% more likely to complete it compared to the old one. The onboarding wizard keeps converting at a higher rate than the regular shop.

Customers acquired through the onboarding wizard are likely to sign up at a higher frequency than those acquired through the shop or other routes, with a high percentage of users leaving their email to keep up with the brand.

The flow is fully accessible and can be completed with keyboard inputs, testing new copy and alternative questions is easier than ever thanks to optimize.

We keep working to improve the onboarding wizard and add more capabilities to it to make it the perfect meal recommendation tool.

You can find the onboarding wizard at the allplants website.

main takeaways, allplants

"When working in a start up your resources can vary and you have to be ready to make do with what is available at the moment"

"Priorities can switch at any moment! Be ready to pivot"