KLOA iOS Keyboard Usability Study
Evaluating the learnability of a novel gesture-based iOS keyboard through exploratory testing, structured usability research, progressive onboarding, and synchronized behavioral observation.
As part of my master's program in Human Centered Design & Engineering at the University of Washington, I worked on a four-person team conducting a usability study for KLOA, an early-stage startup developing a novel third-party iOS keyboard.
Unlike a conventional keyboard, KLOA used combinations of taps, double taps, and swipes within different touch regions. The interaction model could potentially provide an alternative way to type, but first users had to understand how it worked and why they would want to use it.
Our initial exploratory testing quickly revealed a fundamental problem: unfamiliar users struggled to understand the keyboard well enough for us to meaningfully evaluate more advanced use.
Rather than proceeding with the original study approach, we used those findings to redesign the research itself. We created progressive onboarding materials, built a purpose-designed testing environment, and developed a synchronized recording system that let us observe the interface, the participant's physical interaction, and their reactions together. The resulting study helped distinguish between whether the interaction model could be learned and whether users would be willing to overcome the initial learning barrier to adopt it.
I worked across research planning, client communication, testing infrastructure, observation, analysis, and presentation.
- Worked directly with KLOA's founder and developer to understand the product and research goals
- Helped define the research questions and usability study approach
- Participated in early exploratory testing of the keyboard
- Helped the team revise the research plan based on those early findings
- Analyzed and documented the keyboard's modes, gestures, and interaction model
- Created visual onboarding materials explaining unfamiliar interactions progressively
- Designed and built the physical and technical recording setup for formal testing
- Combined three simultaneous video sources into synchronized composite recordings
- Used synchronized timecode to measure task and interaction completion times
- Participated in formal usability testing and analysis
- Helped develop findings and recommendations
- Contributed substantially to organizing, editing, and visually refining the final report
- Presented findings and recommendations to KLOA's founder and the UW HCDE class
KLOA was attempting to introduce a fundamentally different way of typing on an iPhone.
Instead of presenting a familiar keyboard layout and asking users to learn incremental improvements, the product relied on a new set of touch regions and gestures. Users needed to understand where interactions occurred, which gesture to perform, what that gesture produced, and how the different interactions fit together.
Even our research team initially found the keyboard difficult to understand. Only one member of the team was able to use it successfully without substantial difficulty, and not quickly.
That created an important research problem.
If participants could not get far enough into the interaction model to understand it, a formal study might tell us little more than “people don't understand this yet.”
We needed to determine whether the underlying interaction model remained difficult after appropriate onboarding, where specific problems occurred, and whether users could become more capable with experience.
Before building a larger formal study, we conducted approximately 10 informal tests by handing phones with the keyboard installed to people and asking them to try it. The response was immediate. Participants generally became frustrated and returned the phone within roughly five minutes. Those sessions showed us that simply putting the keyboard in front of participants would not produce the research we needed. People were struggling too early in the experience to meaningfully evaluate the broader interaction model. Instead of treating that as the conclusion, we treated it as a finding that needed to change our research approach.
Before we could teach or evaluate KLOA effectively, we needed to understand it well enough to explain it clearly. I worked through the different touch regions, taps, double taps, swipes, modes, and resulting input behaviors and helped map the interaction model. This took substantial effort. The difficulty we had describing the system became useful evidence in itself: if we could not explain an interaction clearly, asking a new participant to discover it independently was unlikely to produce useful insight into its longer-term learnability.
As a team, we developed an onboarding approach that introduced interactions progressively rather than exposing the entire system at once. I created large visual onboarding materials that we placed beside participants during testing. Each interaction was introduced as the participant reached the corresponding portion of the study, allowing them to learn the model incrementally. This let participants get substantially further than they had during the initial informal tests and allowed us to investigate where difficulty persisted after instruction. The onboarding did not eliminate the learning challenge. It gave us a way to study it.
The interaction was difficult to analyze from a single recording. A direct screen capture could show exactly what happened in the interface, but it could not show how the participant physically approached the phone. A camera looking at the device could reveal their hands and gestures, but their fingers could obscure the screen at the exact moment an interaction occurred. A participant-facing camera provided another view into hesitation, frustration, and reaction. I designed and built a recording system that combined three simultaneous feeds: • Direct recording of the iPhone screen • A physical view of the participant interacting with the phone • A view of the participant's face and reactions I adapted technology I already used for live visual performance, using Isadora by TroikaTronix to combine the feeds into a synchronized composite video in real time. Each view provided different evidence. The screen recording showed whether an input registered and what happened in the interface. The physical view showed whether participants were confidently performing an interaction, repeatedly attempting it, hesitating, or hunting and pecking. The participant view provided additional context around confusion and reaction. Having all three views synchronized made it possible to move between these forms of evidence without later reconstructing and synchronizing independent recordings.
The synchronized recordings also gave us a common time reference for analysis. I reviewed the recordings and captured the time participants took to complete different steps and interactions. This helped us compare where participants moved relatively easily and where complexity or unfamiliar gestures caused them to slow down. The setup therefore supported both qualitative observation and behavioral timing rather than relying entirely on participant comments or researcher notes.
With onboarding and recording infrastructure in place, we conducted the formal study with six participants using a consistent script and sequence of tasks. The results showed a more nuanced picture than the initial informal tests. 6 of 6 participants initially misunderstood the input model. All participants improved as they progressed through the study. By the end, 5 of 6 participants could describe the full set of interactions, indicating that the system could be learned with instruction and practice. At the same time, errors increased as interactions became more complex, and participants continued to show low confidence and substantial friction. The distinction was important: the product was not necessarily impossible to learn, but its initial learning cost represented a serious barrier to adoption.
Our recommendations focused not only on individual usability problems but on the larger adoption challenge. We recommended stronger guided onboarding, clearer communication of the product's value, and ways to reduce the cognitive load required before users could become productive. For a novel input method, users needed a compelling reason to invest in learning an unfamiliar interaction model. Improving individual gestures alone would not solve that larger problem. We compiled the findings, observations, quantitative results, video evidence, and recommendations into a detailed client report and presentation.
Findings Report Delivered to Client
The initial exploratory sessions demonstrated that participants could not get far enough into the product for the planned study to answer the questions we cared about. Rather than simply documenting that failure, we changed the study design by introducing structured onboarding and more detailed behavioral observation.
Providing onboarding could potentially influence usability results, but completely unassisted use was preventing us from studying the deeper interaction model at all. Progressive instruction let us ask a more useful question: Once people understand what they are supposed to do, can they learn and use the interaction model?
No single camera angle provided enough information to interpret participant behavior confidently. Combining direct screen capture, physical interaction, and participant reaction made ambiguous moments easier to understand and gave us a richer record for later analysis.
The synchronized recording setup allowed us to examine completion time and behavioral difficulty alongside comments and observations. This helped identify where increasing interaction complexity translated into measurable difficulty.
The study suggested that participants could become more capable with the keyboard over time. That did not automatically mean they would choose to invest the effort required to learn it. Our recommendations therefore addressed both usability and the need to communicate enough value to justify the learning curve.
We presented our findings and recommendations directly to KLOA's founder and delivered a detailed research report containing the study methodology, observations, results, and recommendations.
The research provided evidence that the interaction model could become more understandable with onboarding and practice, while also identifying significant barriers around initial comprehension, complexity, confidence, and the product's value proposition.
KLOA's founder continued developing the keyboard after our engagement, and I later saw a subsequent internal version of the product. Because I was no longer involved in its development, I do not have enough visibility to attribute specific later product changes to our recommendations.
This project demonstrated the value of using early research to evaluate the research plan itself. Our initial informal testing showed that proceeding directly into a formal study would mostly reproduce a problem we already understood. Changing the study allowed us to investigate why the problem occurred and whether users could move beyond it.
It also reinforced the importance of distinguishing learnability from adoption. A novel interaction can become usable with instruction while still asking too much of users unless its benefits are clear enough to justify the effort.
The project also showed me how useful methods from outside traditional UX practice can be when adapted to a research problem. Technology I already knew from live visual performance allowed me to build a synchronized observation system that captured interface behavior, physical interaction, and participant response together.
Finally, it reinforced the value of clear research communication. Complex findings become much more useful to a client when the report, visual evidence, measurements, and recommendations form a coherent story rather than simply documenting everything the research team observed.