SFX (SmashflyX)

Context

SmashflyX (SFX) was an initiative to unify two enterprise recruitment marketing platforms into a single, cohesive experience following Symphony Talent’s acquisition of Smashfly.

The effort focused on rethinking workflows, information architecture, and feature behavior across both systems to reduce fragmentation while supporting existing user mental models. Given the scale and complexity of enterprise recruiting, early discovery was critical to avoid introducing friction during migration and adoption.

Research Focus

Research centered on understanding who was using the platform, how they worked day-to-day, and how search and filtering supported their core responsibilities.

Although the team had a general understanding of the audience, deeper insight was needed into:

• User goals and success metrics
• Behavioral patterns during search
• How legacy workflows influenced expectations
• Where complexity created friction or workarounds

User Archetypes

Based on interviews with 20 users, three primary role-based archetypes emerged:

Sourcer

Specializes in identifying and engaging candidates to build active pipelines.

• High engagement
• Heavy reliance on search and filtering
• Primary user of advanced search capabilities

Recruiter

Manages candidates throughout the hiring lifecycle.

• Moderate engagement
• Uses search primarily within individual profiles
• Focused on evaluation rather than broad discovery

Talent Marketer

Responsible for employer branding and candidate attraction.

• High engagement
• Strong focus on campaign performance and reach
• Uses search as a supporting, not primary, tool

These distinctions were critical in understanding who search needed to serve first—and which behaviors represented edge cases.

Research Methods

Research combined qualitative and behavioral approaches:

• Pre-interview surveys to understand search frequency and intent
• 1:1 contextual interviews with task-based observation
• Legacy platform walkthroughs to identify inherited behaviors
• Comparative analysis between platforms

Rather than relying on self-reported behavior, users were asked to perform real search tasks while their actions were observed and documented.

Key Findings

Across 11 observed sessions:

• Users averaged 4.2 steps before reaching intended results
• 60% of users applied filters after an initial search
• 90% of users consistently relied on 4–5 filters
• Only one user accessed more than five filters
• All users initiated search from existing candidate lists rather than from scratch

Primary pain points included:

• Boolean logic was difficult to understand and error-prone
• Filter labeling lacked clarity
• Navigating, refining, and saving searches was cumbersome
• Search results often surfaced unrelated or irrelevant profiles

These findings challenged assumptions about the need for highly complex filtering interfaces.

How Research Shaped Direction

Research indicated that progressive disclosure was more appropriate than exposing all filtering options at once.

• Prioritize the most-used filters while maintaining access to advanced options
• Separate advanced filtering into an expandable, secondary view

Rather than optimizing for edge cases, the direction emphasized supporting the majority workflow while acknowledging complexity where needed.

Outcomes

While not all research recommendations were fully implemented, the work produced meaningful outcomes:

• Clarified which behaviors were core vs. edge cases
• Informed decisions around progressive disclosure in search
• Reduced ambiguity around search complexity during platform unification
• Helped de-risk early design concepts prior to deeper investment
• Provided a shared understanding across product, design, and engineering

Reflection

This project reinforced the role of research as a risk-reduction tool, particularly in the context of platform unification and acquisition.

While business priorities ultimately influenced which recommendations moved forward, the research helped make tradeoffs explicit and grounded decisions in real user behavior.