Self-Service Checkout

Chewse | 2017–18

Allowing new customers to order first meals online

Chewse is a corporate catering service that offers family-style meals. The primary user would most often be an office manager who was responsible for choosing a meal service for company lunches or specific one-off occasions.

This project aims to improve the first-time order experience for new customers.


Colin Fike

Justin Lee

Kevin Li
VP of Product

Pierre Calzadilla
Head of Growth

Rustin Coburn
Marketing Manager


The process for an inbound customer to get their first meal was heavily manual and involved a lot of offline communication.

Historically, inbound visitors would have to leave their contact info on a form, and Chewse would have a sales representative reach out. Once initial contact was made, there’d be a phone conversation and the sales representative might set up a series of trial meals.

Chewse’s main digital product consisted of a client-facing Meal Calendar that showed upcoming meals. However, we determined there was an opportunity to set up systems for scale by automating that first order and empowering first-time customers to order their own meal.


The broader goal of the project is to supplement our existing high-touch sales process with a self-service model. The sales team had historically closed accounts every month on the order of the low teens. We wanted to experiment with growth methods that could potentially be an order of magnitude greater.

The goal was to validate self-service in the wild, and that involved getting an initial version live as soon as possible.

In concrete terms, we wanted to achieve the following:

• Design a feasible initial solution within two weeks
• Ship within another four weeks
• Continue to aggressively iterate for another four weeks
• Reevaluate

The initial topline metric we aimed for was 100 first orders in the first four weeks between two cities: LA and SF.

My Role

I designed the initial version we shipped and created new collateral alongside the Marketing team to help drive traffic and support the content strategy.

After the launch, I actively designed further iterations of the experience on a week-to-week basis, conducting usability tests and interviews to cross-reference against the metrics we were seeing online.

I currently collaborate with various members of the leadership team to actively define new experiments and test hypotheses.

Understanding the customer

Office managers

At a glance:
- Majority professional women in their mid-20s to mid-30s
- Motivated to keep their team happy
- Keenly aware of team members with unique dietary restrictions or preferences
- Given a budget from company
- Juggles many different services and online tools

Chewse doesn’t position its main value proposition as the actual food. The main value comes from a team that’s energized and unified as a result of eating together — whether at lunch or events organized specifically for employee bonding. Office managers are the people at an organization who own planning and arranging for team and company events, on top of several other functions that vary by company.

These insights were distilled from past interviews and research sessions I’ve had with specific office managers.

Initial User Flow

Initial Prototype

Much of what colored the design decision-making was made from working backwards from our timeline. Considering that we budgeted two 2-week sprints for engineers to get to a launchable MVP, I involved them early in the design process to check where certain bottlenecks might be in the implementation process.

Technical ease of implementation sometimes took priority over best end experience for the user, with the understanding that since the whole self-service checkout was essentially an experiment, getting it live and collecting real information was the highest priority.

Below are some comments on a few of the trade-offs considered in the main screens of the user flow.

Quick note on mobile: Due to pre-existing traffic data, the designs were focused less on the mobile experience for the initial version.

First Pass: InVision


Trade-offs considered
- Growth wanted lead form up front
- First page of pre-existing sign-up flow
- Reduce friction to get to Select Meal step

Step was soon nixed after the launch.


Trade-offs considered
- Wanted to collect information for most common types of recurring meals upfront (weekly, monthly)
- Based off a pre-existing user flow of collecting schedule information
- Bigger concern for self-service checkout was just the date and time of the very first meal
- Engineering effort for schedule input would have to create new custom components

Step was soon nixed after the launch as well.

Select Meal

The Select Meal step is the primary focus of the self-service flow. The following screens are some of the explorations that went into the initial pass.

This concept of a carousel with an expanding selected card was dropped due to clunky interaction.

For the initial launch, a carousel with modal of meal details was the preferred route due to ease of implementation. However, the trade-off we settled on was not having a dedicated page for a specific meal.

Also considered was displaying the meals in a grid. However, we knew that we wouldn’t have a large number of meals available for our initial launch, and we settled on a carousel to keep the page shorter and impactful. We end up revisiting the grid shortly after the launch.

A dedicated meal page would be revisited for future iterations.

Various explorations of laying out meal details.

Delivery Info

First version of the delivery info step involved breaking up a small questionnaire into three distinct steps.

However, we decided to consolidate it into one view to expedite the process. We flagged this screen as a potential drop-off point. There was an inherent tension in ensuring we collect enough information for our in-house delivery drivers and making sure the user didn‘t feel overwhelmed with questions.

Payment and Account Creation

The first pass involved coupling the Payment and Account Creation steps together. However, a hypothesis I have is that users might not want to commit to creating an entire new account to order a meal.

Due to timeline, we decided to stick to this version, and revisit the step once we had enough traffic going through the flow.


Initial Metrics

Following the launch of the self-service funnel, we saw a large number of people bouncing at the first prequalify step. We decided to address it quickly — to learn from data as much as possible, we had to make it users to get to the Select Meal step.

Revised User Flow



Immediate Changes

We cut out the first prequalify and schedule steps to reduce the friction for users to get to the Select Meal step.

However, to show the proper meals for each metro area, we still needed to ask for the zip code information, so we decided to test a lightweight modal.

Marketing Experiment

Content messaging and brand adjustments

We considered the self-service checkout holistically, and that includes the very top of the funnel which includes ads and the landing pages. I collaborated with marketing on messaging and brand content experiments to further qualify the people that wound up in the self-service funnel.

Landing page concepts we conducted user tests with.

Clarifying offerings

Due to the compressed timeline, we ended up doing usability tests in parallel after the initial version of the flow was live. We had suspicions that the carousel was too subtle and reducing the impact of our curated selection of meals.

We worked with the Restaurants team to pull together more meals, and we decided to test the grid layout and break up the offerings into organized buckets.

Also, while the majority of our existing users visit the site on desktop, the majority of new traffic came from mobile, so we considered the Select Meal page from the group up with mobile in mind.

Looking ahead

We can think of targeting improvements to the self-service project in two large areas: marketing and the product flow. Top of the funnel prequalifying helps send the right people to the funnel to begin with, and once the right people land, then the problem can focus more on smaller optimizations.

However, the raw traffic through the funnel hasn‘t been substantial enough for us to learn about the actual flow, so we have started to supplement the queued improvements to the product flow with usability testing.

A lot of the challenges about this project came from a focus on speed, but as a result, the rate of learning has also been greater.

The engineers are currently working on implementing the revised grid of offerings, and the plan moving forward is to continue testing smaller experiments in parallel with larger user flow improvements. Promotional incentives and shorter lead times for ordering are the next operational levers we are playing with.