TemuKerja — Reducing Interaction Demands Across the Job-Search Journey

Digital job portals make opportunities easier to access, but access does not necessarily make the experience manageable. Novice job seekers must interpret whether a role fits their situation, repeatedly prepare information, navigate different application processes, and then understand what happens afterward.

Unclear information, repetitive actions, limited feedback, and a lack of control can add demands to an already difficult journey.

The research began with 8 novice job seekers through contextual observation and semi-structured interviews. Participants used the portal they normally relied on and searched for a vacancy they considered suitable.

I observed how they searched and filtered, what information they used to evaluate suitability, where friction appeared, how they responded when support was insufficient, and what happened beyond simply finding a vacancy. Interviews then explored previous experiences, difficult activities, information needs, expected support, and responses to frustrating interactions.

The qualitative findings shaped a broader questionnaire. 114 responses were collected, with 105 valid responses from people who had submitted through a job portal in the previous six months. The survey prioritized recurring problems rather than attempting to statistically represent all job seekers.

Among valid respondents, 70.5% were actively searching for employment, 98.1% had less than two years of professional experience, and 66.7% most frequently used desktop devices. These characteristics supported both the novice-user focus and the decision to develop a desktop prototype.

Because this was an academic research project, the design criteria needed to go beyond conventional ideas of a “good UX.” Two theoretical perspectives helped interpret the findings and translate them into design qualities.

The frameworks were not academic decoration. Together with the research, they defined what the experience needed to achieve and later determined how the prototype would be evaluated—connecting research, design decisions, and validation.

Satisfaction did not mean making job searching inherently enjoyable. It meant ensuring the product worked according to users’ needs without adding avoidable barriers of its own.

The findings did not automatically mean another job portal was the answer. I compared three product models before committing to a direction.

The three problems were connected. Poor information made evaluation harder; applying introduced repeated administrative work; and unclear progress created uncertainty after users had already invested that effort. The product therefore preserved context across exploration, evaluation, application, and tracking instead of treating each stage as an isolated workflow.

I first confirmed the low-fidelity prototype with 5 target users in moderated think-aloud sessions. The purpose was not to measure final usability, but to determine whether the information hierarchy, labels, interaction sequence, interface states, and data-reuse model were understandable enough to justify moving forward.

Findings led to revisions in information hierarchy, labels, transitions, and how reused profile information was communicated before I moved into high-fidelity prototyping.

The high-fidelity prototype was tested through 3 formative iterations with 5 participants per iteration. Every round evaluated vacancy exploration, application using stored information, and application progress monitoring.

Evaluation combined task success, direct versus indirect completion, time on task, Single Ease Question, adapted NASA-TLX workload, and the Usefulness, Ease of Learning, and Satisfaction dimensions from USE. Observations, interaction paths, comments, and post-test interviews helped explain the quantitative results.

The adapted NASA-TLX was used only as an indicator of subjective interaction workload—not as a measure of overall psychological job-search stress.

Cross-iteration results were treated as formative evidence rather than a controlled experiment. Iterations 2 and 3 offered the clearest descriptive comparison because they used the same participants and testing platform.

The prototype demonstrated the interaction concept, but several parts depend on systems beyond the interface. Reusable applications require structured storage for profile data, documents, responses, and the exact version submitted. Tracking requires recruiters, applicant-tracking systems, or other recruitment systems to provide reliable updates. Company information and reviews require validation, moderation, and governance.

An initial product could support profile and document management, in-platform applications, internally managed status histories, attribute-based recommendations, and administrator-managed company information. Automated CV extraction, external ATS integration, and user-generated reviews would introduce additional technical and organizational dependencies.

The final prototype performed strongly in formative usability testing, but the evidence describes interaction quality while using the prototype. The adapted NASA-TLX result reflects perceived task demands during interaction; it does not demonstrate reduced psychological stress associated with job searching as a whole.

Because this was an interactive prototype rather than an operational recruitment platform, it does not establish effects on employment outcomes, application conversion, long-term behavior, recruitment success, data accuracy, production performance, or business metrics.

Competition, employer decisions, rejection, personal circumstances, and broader uncertainty cannot be removed by an interface. The claim is narrower: interaction design can reduce avoidable demands introduced by the digital product itself.

This project showed me that solving an interaction problem is not the same as adding the feature that appears to correspond to it. More filters did not automatically improve exploration. Reusable information still needed visibility and control. Tracking only became useful when it appeared where users expected it.

It also changed how I evaluate design. Task success was already 100% in the first iteration. Directness, time, workload, perceived ease, and observed behavior revealed the interaction problems that completion alone concealed.

Scan the QR code to access the supporting academic materials from this project, or click here.

Included materials

• 01Undergraduate thesis
• 02Research poster
• 03Presentation slides