Technology
Technology Behind FlowScan
ConvertScale combines automated signal detection, algorithmic scoring, and LLM-powered interpretation to diagnose funnel performance at scale. This page explains the four-stage diagnostic pipeline that powers every FlowScan.
Our approach
Diagnosis before optimization. Always.
The conversion optimization industry is built on tools that measure what happened — clicks, views, heatmaps, session recordings. But measurement is not diagnosis. Knowing that visitors leave is not the same as knowing why.
ConvertScale takes a different approach. We combine automated signal extraction with rule-based scoring, and use a large language model to explain the findings and write the recommendations.
The result is not a dashboard. It is a diagnostic report with a FlowScore, severity-ranked issues, and plain-English recommendations written with AI.
Diagnostic architecture
Four stages. One structured output.
The FlowScan diagnostic engine operates in four sequential stages — each building on the output of the previous one to produce a comprehensive, actionable report.
Collecting diagnostic inputs
Signal Detection
The first stage captures raw signals from the funnel — page structure, content elements, load behavior, CTA positioning, trust markers, and technical metadata. These are the observable inputs that feed the diagnostic engine.
Identifying specific problems
Issue Detection Engine
Detected signals are evaluated against our diagnostic framework to identify specific, categorized issues. Each issue is mapped to one of nine scored categories and assigned a severity level.
Quantifying funnel health
Structured Scoring Model
Issue findings are aggregated into category-level scores and a composite FlowScore. The scoring model weights issues by severity and gives each category a fixed weight in the total.
Large language model analysis
AI Interpretation Layer
Structured findings are processed through cloud-hosted large language models that generate natural-language explanations, cross-reference analysis, and implementation-ready recommendations for each issue.
Nine scored categories
The framework behind every FlowScan.
Every funnel is evaluated against the same nine categories — creating a consistent, structured, and comparable diagnostic framework.
Conversion Architecture
How page structure, calls to action and lead capture guide visitors toward the conversion step.
Tracking & Data
Analytics tools, tracking pixels and event setup detected on the page.
UX Friction
Usability problems such as popups, a broken mobile layout and poor readability.
Visual Design
Design quality and branding consistency, such as outdated design and low-quality images.
Trust & Credibility
Proof that makes the offer believable, such as testimonials, reviews and client logos.
Performance
Load speed and page weight, measured for mobile.
Security & Technical
Technical hygiene such as mixed content, security headers, broken links and script errors.
Marketing Intelligence
Whether the funnel is ready for paid traffic: ad readiness, a retargeting pixel and audience segmentation.
Automation & Follow-Up
What happens after sign-up, where the scan can test it: thank-you page, confirmation and next step.
Funnel intelligence layer
Building Funnel Intelligence.
FlowScan is the diagnostic engine. Every report is written from that scan's own findings and built to be actionable.
Pattern Recognition
Findings come from fixed, rule-based checks applied to each scan. The rules do not change based on past scans.
Benchmark Data
Once enough real scans exist, a FlowScore is compared against a cohort of prior scan scores. Below the minimum cohort size, no benchmark is shown.
Recommendation Engine
A language model writes each scan's recommendations from that scan's own findings.
Diagnostic Memory
Each scan is stored as its own result, so a follow-up scan can be run after changes are made.
Infrastructure philosophy
Three principles that guide our technology.
Systematic diagnostics
Funnel problems are detected through structured diagnostic analysis rather than manual audits. Rule-based findings are reproducible; AI-judged checks can vary slightly between scans. Every finding is explainable and traceable to a concrete diagnostic input.
Scalable intelligence
The platform is designed to run many scans at scale, with each scan stored as a structured diagnostic result.
Actionable insights
Diagnostics are translated into clear recommendations that help improve funnel performance. Findings are specific enough to implement — not vague enough to ignore.
Infrastructure
Built to run browser scans at scale.
Every FlowScan requires significant compute — from headless-browser rendering to language-model calls. Our infrastructure is designed to scale diagnostic workloads across cloud environments.
LLM Integration
The AI interpretation layer runs on cloud-hosted LLM infrastructure, using separate prompts for page analysis, visual judgement and recommendations to generate diagnostic summaries and implementation-ready recommendations.
Async Processing Pipeline
Distributed task workers handle browser rendering, signal extraction, and language-model calls. Built on Celery with Redis for queue management, enabling concurrent scan processing at scale.
Structured Data Store
PostgreSQL stores structured diagnostic output: scan results, category scores, issue catalogs, and the written recommendations.
Compute Requirements
Each scan involves headless browser rendering, DOM analysis, multiple LLM inference calls, and report generation. As scan volume scales, cloud infrastructure becomes the critical growth bottleneck — managed databases, auto-scaling worker pools, LLM API throughput, and object storage for diagnostic artifacts.
Built for FlowScan
The technology powers the product.
Everything described on this page is the engine behind FlowScan — ConvertScale's first product. When you run a FlowScan, the full pipeline scans your funnel, calculates your FlowScore, and produces your diagnostic report.
Signal detection, issue classification, structured scoring, and LLM-powered interpretation — they all come together in a single, automated diagnostic experience.
See FlowScan in actionDiagnostic Pipeline
Built with
See it in action
Experience the diagnostic engine firsthand.
Run a FlowScan on your funnel and see exactly how our technology identifies issues and generates actionable intelligence.