Automating Daily Menu Design
Can research improve user experience through optimizing operations?
Overview
For a meal subscription delivery company with hundreds of items, 50+ meals every day, and multiple categories and diet plans, menu design and generation is a long and tedious task. Multiple things go into consideration: meal ratings, meal composition, the status of the supply chain and what they're able to get and seasonal and new meals. It was also a manual process, and therefore, not scalable. Identifying that the best course of action is to automate this process, this research tackles do we get there.
Problem Statement
Meal subscription menus are more complex than restaurant menus. They require a balance between new and popular items, catering to both adventurous eaters and picky ones. With over 50 daily menu items, the process of generating menus has been successful, consistently receiving high ratings and minimal complaints about variety. However, the process is currently entirely manual and requires careful curation by the responsible individual, adhering to tried-and-tested Food R&D constraints. The problem is that this process has never been documented, making it difficult to expand the menu or automate the process. The question remains: how can we automate it?
Impact on user experience:
Automation allows the Food R&D team to focus on research and development
A unified, well-documented process makes it easier to trace feedback, isolate the issue, and make improvements
Reducing human error allows for a consistent pleasant experience.
Kick-off Meeting
This initial meeting serves as a platform to create a shared understanding of research goals and scope.
During this meeting, the product owner went over the following:
Orienting the team on the task at hand.
Breaking down the steps
Takeaways:
Setting success metrics
Understand the technical limitations
De-risking measures
User Interview - Subject Matter Experts
Understanding the manual menu generation process is crucial to building the foundation to automate it. The desired outcome of the interviews is complete documentation to build the foundation to automate the process.
A semi-structured contextual interview was conducted. The expert is the Food Operation Specialist overseeing the task of menu generation.
The outcome of this interview was a comprehensive decision tree with all the possible combinations to choose a specific meal for a slot on the menu.
Note: the decision tree nodes are redacted
Solution Workshop with the Team
Agenda
After the insights from the experts were turned into a decision tree, the team was gathered to workshop the best automation solution.
Participants
The Operation Squad → PM | Engineers | UX Designer
Food R&D → Food Ops Team
Technical limitations uncovered
The criteria the food team uses to design the menu are not the same criteria saved in the meals database.
Workaround
Import the data from the database to an external sheet
Re-write the criteria from the food team to meet the database identifier
Example
Before the workaround
select meal WHERE: type = breakfast, & diet = dietidentifier, & base = parfait OR pancake OR oats, & taste = sweet
After the workaround
select meal WHERE: type = breakfast, & diet = dietidentifier, & taste = sweet
First Iteration Release Plan
Release in the smallest market first
Allow the Food Ops team to make as many changes as necessary to the menu generated based on the commands
Document all changes and reasons for the change
Monitor any meals’ ratings changes and variations
Add the necessary identifiers generated from the changes the Food Ops team to the database
Make the changes necessary for the second iteration