Establishing Use Cases for Sequence

UX Research Case Study Portfolio – Mabel Tan

Explorative: Establishing use cases

Sequence: I was a co-founder working on product and user research at Sequence. We proudly ran as a bootstrapped startup. Sequence was an outsourcing service for image annotation and data labelling for machine learning projects.

Research Goal:

To explore the workflow of data scientists and discover areas where our proposed solution would be most useful.

Method: User interviews, Surveys

We recruited a total of 15 participants who were data scientists by profession. User interviews were conducted remotely over Skype.

Participants were also asked to complete a qualitative questionnaire regarding their general activities at work. The questionnaire was followed up with a 30 minute interview

over Skype. The user interviews asked open-ended questions regarding their workflow, methods, and pain- points.


From the the user interviews, we uncovered generally two broad personas.

Persona A: Mathieu

Works primarily with machine learning models that deals with image recognition and data labelling at a medical science organisation.

Pain points: Acquiring clean data at a high enough volume. Annotating and labelling data himself. The highly competitive nature of the medical industry means restrictive NDA’s often prevents him from outsourcing the tedious task of annotating images.

Persona B: Elaine

Works primarily with machine learning models that deals with raw log data. She works at giant multinational tech startup.

Pain points: Business Analysts don’t normally understand her scope of work well enough – which results in briefs that are not suited for machine learning models. Munging data.


  • The way data is annotated, labelled and categorised depends highly on the nature of the data.
  • Image annotation and data labelling tasks are often simple to execute but extremely tedious to do at high volumes.
  • Marketplace sites like Mechanical Turk is a cheap way to crowd-source data labelling. However, data scientists often find that the task of managing crowd-workers eats up time that they might have preferred to spend on other tasks.


This gave us the confidence to explore our project further with Persona A: Mathieu in mind. As we were able to see in which use case we could intercept the workflow and provide a solution that would counter the pain points. We proceeded with this information as implications for design and produced mock-ups which eventually took us to our Beta version of our platform. We used the insight around NDAs, privacy, and confidentiality to shape our business development strategy and processes – around a highly specialised service that was secure and trustworthy.