From 3.7% to 8.2%: analytics system and conversion growth in Facebook over 2 months

Stages of building voice and end-to-end analytics system

• Stage 1. Tool selection and integrations

  We conducted an audit of voice analytics, end-to-end analytics, and automation tools. We compared off-the-shelf SaaS solutions with a custom no-code stack integrated with Facebook. Ultimately, we chose a custom setup (RingCentral, ChatGPT, Zapier, Albato) due to flexibility, cost efficiency, and faster implementation.

• Stage 2. Prompt development and iteration

  We built an AI system for analyzing sales calls, gradually increasing prompt complexity. Each version added new evaluation layers, lead qualification logic, and structured recommendations. Prompts were continuously refined based on real calls and feedback, improving accuracy over time.

• Stage 3. Launch and sales manager control

  We implemented daily call evaluation with AI feedback delivered automatically via Telegram. This helped identify weak points in the sales manager’s performance and gradually improved scores from 1–2 to 9+. Improvements were supported by ongoing script refinement and communication optimization.

• Stage 4. End-to-end analytics implementation

  We built a system tracking the full funnel from Facebook ad click to signed contract using GA4, Typeform, CRM, and call tracking. This provided full funnel transparency and enabled campaign performance analysis and bottleneck detection. Based on the data, multiple marketing hypotheses were tested.

• Stage 5. Automation layer

  We automated deal creation, call processing, CRM updates, and document generation using no-code tools. This eliminated manual work for the sales manager and reduced lead processing time to seconds. All core processes became fully automated and scalable.