Research, Strategy, Concept, Testing
SEEK
2020
SEEK is Australia's largest and most widely utilized job search site. The notifications product suite spans all customer touch points from onboarding and applying and role management, as well as engagement throughout their lifecycle.
A key product SEEK offers users is an algorithm engine generated list of recommendations for roles that suit the candidate profile. It forms an integral part of SEEK’s communication strategy to Candidates. The volume of the role recommendations is greater than any other communications product, creating opportunity for user engagement, and the existing product did not address this potential.
We turned to both quantitative and qualitative data sources to get the complete picture. While our data analytics showed that the product was performing well in terms of opens, click throughs and application engagement, survey responses and insights gleaned from continuous discovery interviews told another story. Our users expressed frustration that their role recommendations are inaccurate or irrelevant, often missing the mark.
How can we engage our users to provide additional information that will improve the accuracy of our role recommendations algorithm and provide a more personal experience?
Designed a deep dive on our customer segmentation and developed hypotheses for how we might personalize and create engaging content, validate whether existing internal subgroups are still applicable to real users, and investigate the behavior patterns that could cause errors in AI matching
Created clickable prototypes of several concepts based on a "give to get model" of providing tangible value to drive engagement, based on how each group might respond differently to determine whether this might be a viable strategy for gathering better datasets to improve algorithm accuracy and give us more robust insights to explore future product potential
In hybrid interview and remote testing sessions, we asked candidates to walk us through their general job searching behaviors and observed interaction with our product while gathering real time feedback and validation for our on our randomized prototypes
What Was Tested
We utilized a color-coded system to synthesize insights and find patterns from participant sessions to inform our discovery deck of product learnings and next steps
Users have different needs depending on jobseeker activity, more-so than SAT/MID/MAX divisions.
Marketing should work more closely with Product to develop alternate strategies for segmentation and contextual promotional content within transactional email products.
Focus on capturing unique preferences to allow for nuanced seeker situations and improve accuracy of recs algorithm to ensure our relevant recommendations.
While not currently collected on the user profile, gathering this signal would assist with segmentation and serve as a potential profile nudge
Due to unique needs that our product doesn't meet, there is a behavioral preference for self-scouring SEEK as opposed to trust in our suggestions.
Passive candidates use WRU differently, by casually browsing and entering the site via email as opposed to Active candidates who search on platform on a daily- weekly basis and only check email as a catch all.
