Helping breastfeeding mothers assess the efficacy of their sessions


The Project

I helped an early-stage startup working at the intersection of personal health and machine learning prepare for a Wizard of Oz prototype experiment.

Project Details

  • Type: Personal/Side Project
  • Duration: 3 weeks (Nov 2017)
  • Deliverables: iOS app design and evaluation plan for Wizard of Oz prototype

My Role

I designed a mobile app prototype for the WOZ sessions, wrote a WOZ evaluation plan, and worked as a UX design consultant for an awesome team that was eager to create a human-centered product experience.


Gulp helps moms feel more confident about the efficacy of their breastfeeding sessions. It consists of a small wearable device and mobile app. The team had a few goals for this project, and I helped tackle one in particular, to validate this assumption:

Moms will value feedback from a mobile app following a breastfeeding session.

The Challenge

95% of mothers have breastfeeding concerns and 2/3 are not breastfeeding for as long as they’d intended. It can be difficult for mothers to know whether a given session went well or poorly. This often results in frustration or worry.

Gulp Mobile App

One challenging aspect of this short project was that the mobile “app” needed to look real to the WOZ participants.

The Wearable

While I was not involved with the wearable, its role in the experience significantly impacts the mobile app and vice versa. To use, moms place the wearable sensor on their bra, shirt or other convenient location near the baby while she nurses. The rest of the interaction occurs with the mobile app.

Start a Breastfeeding Session

Open the app and tap “start” to begin a breastfeeding session. During the session, Gulp collects audio data from the baby suck, swallow, and breath patterns. Tap “finish” to indicate the end of a session.

Gulp-Start BF Session.png

Reflect and Assess

After the session, the user can fill out a short questionnaire about her experience to help Gulp can generate an accurate result.

Gulp-Assess and Reflect.png

Act with Confidence

To keep results human-friendly tone and understandable, the team started with three: Successful, Moderate, or Concerning. This was an area to pay special attention to in the WOZ. If a result is concerning, Gulp will offer to connect mom with a certified Lactation Consultant to learn more.

Gulp-Act with Confidence

The Journey

Getting up to speed

I first interviewed two co-founders to understand the project and what prior work was done. I picked their brains about the problem space and competitors. 

Afterwards I interviewed the third cofounder and resident subject-matter expert. She helped me learn more about the users and context surrounding the breastfeeding experience.

Do moms value and trust realtime feedback and proactive intervention following a breastfeeding session?

The team’s prior research revealed these insights:

  1. Wearables and other hands-on interactions should be non-interfering to the breastfeeding session
  2. Results from machine-learning and analysis must be translated to be human-understandable
  3. Moms find value in 24/7 access to professional support from lactation consultants

For the WOZ, we knew we couldn’t address #1 – there would simply be interfering aspects of the experience. Instead, we aimed to learn about #2 and #3.

User tasks and quick concept creation

The team identified four user tasks critical to executing the WOZ. The last two were particularly geared at testing our hypotheses.

  • Activate the mobile app – necessary
  • Start, pause, and end a breastfeeding session – necessary
  • Answer a short questionnaire after each breastfeeding session
  • Acknowledge or act upon Gulp’s results of the session
App Flow – WOZ

Refinement and focus

After another round of iteration, we had settled on the following changes:

Elevate the AI aspect

We wanted to see how participants react to this aspect of the technology and experience. Would they trust it?

Add ability to pause and resume the session

We needed this due to technical limitations of the WOZ. Ideally the wearable would detect the start and end of sessions so that moms wouldn’t even need to open the mobile app until they were finished.

Be highly conscious about the density of information

This principle was inspired by a Design for Healthcare event I attended in Seattle. In the context of Gulp, the experience had to balance information availability with information need . If executed poorly, Gulp could feel like extra work or an interference. This risk is magnified by the fact that newborns can nurse 8-12 times per day – that would mean up to the same number of interactions with the mobile app.

To minimize interference with the real world experience, a simple, low-touch, and clean interface design was necessary.

An appropriate personality

The last week was about refining the visual design and planting a few teaser features to gauge participant interest. I facilitated a short design critique of the visual design and we realized it was missing boldness and vitality. It was too much pastel and needed a dose of zest to reflect the energy and excitement of the newborn experience.

Gulp Color Palette Small

App – Quick Evolution

It was important for the mobile app to look convincing and real. With the limited time, we decided that a focus on color would make a substantial impact. To reflect reassurance and comfort, my first palette consisted of a calm light green, soft pink, and airy white.


Writing the WOZ Evaluation Plan

I created a guide to help prepare the team to administer the WOZ study. It contained the following:

  • Defined goals and research questions for the WOZ
  • Suggested methods: semi-structured interviews, diary studies, surveys
  • Scripted interview questions (25), including prompts and probes
  • Measurement: quantitative via Likert scale for some interview and survey questions; otherwise affinity diagramming

Rather than fully script everything out, my goal was to provide the team a few building blocks to conduct and later assess their user research study.



If I had stayed on this project, I would have loved to explore more automation and less UI. For example, in the spirit of minimal interference, are two devices (wearable and mobile app) really necessary? What about trying this UI on a Smartwatch with built-in microphone? Or no UI at all?

Longitudinal Studies

For the WOZ I recommended the team use a simple variation of a diary study. However, for a future design evaluation it would be especially insightful to learn how moms would use the product over a more substantial period of the breastfeeding term.