UX / Product Design Case Study · 2025 · Web App
SocialHub
Designing an AI-powered social media management platform that lets a marketing team plan, create, schedule, measure and collaborate — from a single workspace.
Designing an AI-powered social media management platform that lets a marketing team plan, create, schedule, measure and collaborate — from a single workspace.
Marketing teams manage several brands across Instagram, Facebook, LinkedIn, X and TikTok — usually with a different tool for scheduling, another for analytics, another for design, and a group chat for everything else.
SocialHub brings content creation, an AI assistant, a design editor, scheduling, analytics, team roles and live meetings into one connected product. This case study walks through how I moved from research and information architecture to a validated, shippable interface.
Interviews and a survey surfaced three recurring frustrations that had nothing to do with creativity — and everything to do with the tools.
Switching between a scheduler, an analytics suite, a design app and chat to publish a single campaign. Context is lost with every jump.
Dashboards report what happened but rarely say what to do next — no guidance on gaps, timing, or what to post.
Approvals, roles and hand-offs live outside the tool, so junior staff can't publish without a senior manually reviewing.
I framed the project around four design goals, each tied directly to a pain point uncovered in research.
Plan, create, schedule and report without leaving the product.
Turn analytics into clear, next-step recommendations — timing, gaps, content ideas.
Roles, approvals and tasks so teams can move fast without losing quality control.
A dense product that still feels scannable, learnable and unintimidating.
I diverged to understand the real problem, converged on an information architecture, then diverged again through design before validating with users.
User interviews, survey, competitor teardown of Hootsuite & Buffer.
DiscoverPersona, user stories, journey map, problem & HMW statements.
DefineCard sorting → information architecture → key user flows.
DevelopWireframes, design system, high-fidelity UI across 9 modules.
DevelopUsability testing, iteration, hand-off specs.
DeliverSynthesising interviews into a primary persona kept every later decision anchored to a real person, his goals and his constraints.
Mapping Karl's journey across five stages exposed the emotional low points — the annoyance of weak trend insight and the frustration of tedious setup — and turned each into an opportunity.
The journey's pain points became a focused set of “How might we” questions to guide ideation.
How might we let a team run an entire campaign without leaving one workspace?
How might we turn raw analytics into a clear next action to take today?
How might we make setting up a 10-person team feel guided, not like admin work?
How might we give leaders control over quality while still empowering junior staff?
With a long feature list, I couldn't guess the navigation. I ran a hybrid card sort so real users decided how features belong together — before a single screen was drawn.
Four groups formed consistently. The percentage shows how many participants placed those cards together — the agreement that gave me confidence to lock the structure.
Two cards refused to settle. Participants disagreed on Roles & Permissions (people vs. settings) and Notifications (engagement vs. account). So I kept role management inside Team but surfaced it in Settings too, and pulled Notifications into a persistent global top bar rather than burying it in one section — decisions driven by the data, not a hunch.
The card sort translated almost directly into navigation. Four primary areas, each holding the features users had already grouped — a structure that tested as intuitive because users built it.
Publishing a multi-platform post used to take six tools. The IA let me compress it into one continuous path.
Type a prompt into the AI Assistant.
See live previews per platform, pick a tone.
Adjust in the built-in editor or a template.
Drop into the best-time queue on the calendar.
Track results back on the analytics dashboard.
Electric blue for action, deep navy for focus, generous white space and one accent green reserved for positive signals. Built as tokens and reusable components for consistency at scale.
High-fidelity screens where the research, IA and system come together.
The home screen opens with AI content-strategy insights — content gaps, optimal timing, platform balance — beside the content calendar and scheduled posts. Karl sees what to do next the moment he lands.
Performance trends, AI predictions, audience demographics and competitor benchmarks — each paired with a confidence level and a recommended action, so data always resolves into a decision.
Describe an idea, pick platforms and a tone, and the assistant returns ready-to-post content with live, per-platform previews — LinkedIn, Facebook, X and Instagram at once.
A week grid highlights optimal posting windows so the queue fills itself around when the audience is actually active.
A categorised template library and a layer-based editor mean campaign creative never has to bounce out to another app.
Admin, Manager, Editor and Viewer permissions let leaders delegate confidently — the exact trust problem the research surfaced.
Video meetings, agendas and quick-start join keep hand-offs and reviews inside the product instead of scattered across other apps.
A clean, welcoming sign-in with social auth — the calm, uncluttered promise the rest of the product keeps.
Five moderated sessions on the prototype, task-based. Overall task success hit 92% — and the friction that remained pointed to clear, small fixes.
Users expected to schedule from the post preview, not a separate screen.
The three insight cards read as decoration, so users scrolled past them.
“Editor” vs “Manager” permissions weren't obvious at a glance.
Grounding navigation in card sorting paid off where it matters most: users found things, finished tasks, and did it in fewer steps.
Card sorting removed my own bias from the navigation. The two cards that wouldn't settle taught me more than the ones that agreed — disagreement is a design brief.
An insight the user can't act on is noise. Every AI moment became stronger the second it ended in a button.
A feature-rich B2B tool doesn't have to feel heavy — restraint in type, colour and spacing does the heavy lifting.
The journey map flagged team setup as a low point. A guided onboarding wizard is the natural next iteration.