How 58.com Revamped Its Job Search Page with Personalized Recommendations

01 Introduction

Full‑time recruitment is a core business of 58.com. In autumn 2020 the author participated in a personalized recommendation upgrade for the job search category pages, prompting reflections on design perspective, solution viability, inclusiveness, and fun.

02 Shift from Group to Individual Perspective

With advances in internet and data mining, understanding of users deepens. Previously content was aggregated for groups; now focus shifts to individual perception, job experience, and growth cycle. Insight combines commonalities and diverse traits, forming an abstract decision model for blue‑collar job seekers, confirmed with algorithm and product teams.

The model shows a blue‑collar seeker evaluates options based on multiple decision dimensions and psychological weights. Designers provide a container and rules, replacing the old clustered header with three personalized recommendation zones to meet individual needs.

Zone A – Precise Recommendation

Combines resume, intent, behavior data to present targeted jobs via a “stitched” main and subtitle, configurable across three dimensions; the third dimension is open to operations for flexibility.

Zone B – Expanded Recommendation

Similar stitching, but offers additional, less obvious positions, using an AB combination to balance precision and variety for blue‑collar users.

Zone C – Location‑Based Recommendation

Based on LBS, shows nearby addresses, business districts, or residential areas, allowing users to customize address on a secondary page, enhancing trust.

Additionally, a video‑live job feed is added to meet diverse content preferences.

03 Design Must Consider the Lifecycle, Not Just the Moment

The upgrade represents a “personalization 3.0” era, moving from basic infrastructure to modular, efficient services. The design focuses on long‑term planning, establishing a feedback loop: recommendation content → user → feedback system → recommendation strategy → content.

Feedback mechanisms are enhanced with “you may be looking for”, “not interested”, and a new post‑application tag recommendation to capture intent more accurately.

Feed improvements include three new job card styles, a dialogue‑style card D, reduced lazy‑load count, and strategic placement of quality zones and ads to boost application rates.

04 Design Should Be Inclusive, Friendly, and Fun

The category page balances stakeholder interests, defining business zones and design standards while ensuring extensibility. Visual appeal is heightened with LBS zone cues, “road sign” micro‑animations, and day/night themes, configurable for future weather‑based variations.

05 Conclusion

Personalized recommendation is a mainstream design direction. Future work will deepen user perception layers to improve overall experience. Thanks are given to mentors and contributors, with anticipation of the 3.0 launch in early 2021.