Skip to content
Jon Gifford CXMind

Crowdbotics · 2023–2024

Triage first. Redesign second. Admit what's still unclear.

Scoring a leaky funnel by impact and effort, triaging the highest-impact issues, then relaunching around persona journeys.

Final page designs: the new brand applied to the tested structure, desktop and mobile.
Final Crowdbotics high-fidelity page designs in the new brand style, desktop and mobile

Final page designs: the new brand applied to the tested structure, desktop and mobile.

My role
CX and creative direction lead. Led the audit and impact/effort scoring, journey mapping, taxonomy, creative direction and prototyping; ran client communication.
Team
Delivered as a team with the client; limited in-house development capacity set the build scope.
Constraint
Short timeline and limited in-house development capacity. The first build had to be simplified.
Result
+20% form conversion (relative; 0.5% → ~0.6%) (observed post-launch)
Strategic decision

Phase the work by business impact: triage data and flows now, rebuild taxonomy and positioning next.

  • +20%Form conversion (relative; 0.5% → ~0.6%)Observed post-launch
  • +23%TrafficObserved post-launch
  • 4.25×Crosslink activityObserved post-launch

Conversion is a relative lift on a 0.5% baseline, not 20 percentage points.

Evidence

Crowdbotics builds apps through a platform plus a service layer. The positioning made development look easy, so some prospects dismissed it as too simple. The numbers agreed: 0.5% form conversion, 79% homepage bounce.

We audited the whole experience, including the marketplace and CRM, and scored each area for impact and effort. Three problems kept coming up: redundant navigation labels (Technology, API, Platform), unfinished design language, and content that never said what the company did.

Homepage audit, each issue scored for impact and effort. This is the triage list, not a mood board.
Annotated homepage audit with scored negative findings

Homepage audit, each issue scored for impact and effort. This is the triage list, not a mood board.

Strategic choice

Structure came before style. We rebuilt navigation and taxonomy around persona journeys, then mapped conversion points to those categories so tracking matched real behavior. Low-fidelity wireframes followed tasks across taxonomy levels to find where each persona lost the thread.

Wireframes locating content blades by persona task. High fidelity waited until these flows held.
Low-fidelity wireframes marking target content blade locations across templates

Wireframes locating content blades by persona task. High fidelity waited until these flows held.

Crowdbotics process blade prototype (Figma)

Open in new tab

Tradeoff

Timeline and development capacity meant a simplified first build, with a fuller rollout later. Because the interim fixes carried into the final design, most KPIs improved anyway.

We tracked persona-specific search terms through the funnel so the next round could be judged on behavior rather than opinion. Testing confirmed the personas matched real use. It also showed industry pages drew interest but little engagement, and that content depth fell short almost everywhere. The layout was now good enough that the client started noticing deeper taxonomy problems across the domain — a useful kind of trouble.

Limits

  • 28% of tested users still left unclear what Crowdbotics offers — a messaging problem layout can't fix.
  • Neither round moved the skeptic persona, who arrived to confirm a negative bias rather than decide.
  • Codeshare and marketplace still lacked conversion paths; top-of-funnel content was the biggest open need.