Codifying human expertise into a guided product for SME software discovery

Phase 1:  Market Intelligence

Segmentation Logic: 

• Buyers were categorised into production-led, service-based, and retail/wholesale firms.
• Vendors were grouped into organisation-wide platforms (e.g., ERP, cybersecurity) and finance-specific tools (e.g., billing, payroll).

Competitive Analysis: 

• We audited 17 platforms across UX, filtering depth, recommendation flows, and trust mechanisms. 
• Most followed static catalogue structures with paid listings and lacked contextual guidance.

‍

‍

Phase 2 -  Product Mapping

UX Interviews: We conducted structured interviews with internal consultants, SME buyers, and software vendors to map behavioural roadblocks in the discovery process. Three consistent friction points emerged:

• Buyers could describe problems, not product needs.
• A wrong choice was seen as high-stakes and difficult to reverse.
• Trust in recommendations was low due to concerns around paid listings.

‍

‍

Persona Matrix: We synthesised interview findings into four distinct behavioural archetypes:

• First-time Buyer – cautious, unfamiliar with software categories.
• Over-extended Operator – multitasking, decision-fatigued.
• Skeptical Accountant – risk-averse, focused on compliance.
• Ambitious Owner – growth-driven, tech-forward.

Each persona informed tone, interface design, and copy strategy.

‍

‍

Phase 3: Architecture & Experience Design

Data Architecture: We developed a flexible entity-relationship model to support evolving buyer inputs and scalable recommendation logic. Four key entities formed the foundation:

• Users: Captured role, persona type, and language preferences
• Diagnostic Sessions: Timestamped, schemaless JSON responses for flexible input handling
• Software Products: Tagged by solution category, features, and compliance attributes
• Recommendations: Stored matched products, rationale for selection, and confidence scores

This helped us:

• Map relationships between answers and products
• Update question logic without rewriting the backend
• Keep data flexible for future AI integration

System Design:

We designed the diagnostic flow to mirror how real buyers think

• Each screen was limited to 5 inputs to reduce cognitive load.
• Questions were phrased around real-world challenges (e.g., “Are you struggling to track inventory?” instead of “Do you need an ERP?”).
• Answer formats were optimised for mobile, with clear, tappable choices.
• An “I’m not sure” option was included throughout to reduce friction and dropout.

‍

‍