VS/AI
01Built from real project use
Paid media · analytics · conversion

Growth Intelligence & CRO Engine

Combines advertising, website and visitor-behaviour data to explain what is working, what is wasting money and what should be tested next.

01The problem it solves

Ad dashboards report what changed, but marketers still spend hours connecting platform performance, search intent and landing-page behaviour to explain why it changed.

02What it does
  • Brings advertising and website data into one view.
  • Finds weak campaigns, wasted spending, tired creative and website problems.
  • Explains the likely reason behind a performance change.
  • Creates a prioritized test list and an easy weekly report.
Who built it

Built and consolidated by Vipul Sharma from several versions used for campaign analysis and conversion improvement.

What you give it

Manual input or
automatic data.

Information can be entered by a marketer, uploaded as a file or collected automatically through an approved connection.

01
Advertising dataAutomatic connection or file upload

Campaigns, ads, keywords, spend, leads and sales.

02
Website analyticsAutomatic connection or file upload

Visitors, landing pages, funnel steps and conversions.

03
Visitor behaviourAutomatic connection or export

Clarity recordings, heatmaps or written observations.

04
Business targetsEntered by a marketer

Target cost, margins, location, offer and sales cycle.

What the system checks
  • Cost per customer, cost per lead, return on advertising spend and conversion rate
  • Campaign and ad-group structure
  • Keyword and search-term waste
  • Creative fatigue and frequency
  • Landing-page message match
  • Funnel and form friction
  • Competitor and offer context
How it works

From input
to useful output.

01Ingest→
02Normalize→
03Benchmark→
04Diagnose→
05Prioritize→
06Approve→
07Report
Connectors

API, CSV and export adapters keep live credentials optional.

Model

A common campaign schema aligns metrics across sources.

Diagnostics

Rules and agents test spend, intent, creative and CRO hypotheses.

Evidence

Every recommendation links back to metrics or observed friction.

Delivery

Dashboard, report and experiment backlog remain human-controlled.

What you get back

Files, reports
and next actions.

System outputs
  • Campaign-performance dashboard
  • Customer-cost and lead-cost diagnosis
  • Wasted-spend and search-term report
  • Creative-fatigue alert
  • Landing-page CRO scorecard
  • Prioritized experiment backlog
  • Weekly growth report
What it needs to run

Core tools and
outside connections.

Required connections are needed for the main workflow. Optional connections can be replaced with uploaded files or sample data.

Analysis
  • Python
  • Pandas-compatible data layer
  • Statistical rules
  • LLM evaluation
Application
  • FastAPI-compatible services
  • React dashboard
  • PostgreSQL-compatible storage
Automation
  • Scheduled imports
  • Report generation
  • Approval queue
  • Audit log
Outside connectionIs it needed?What it is used for
Google Ads APIOptional

Live Google campaign and search-term ingestion; CSV mode is available.

Meta Marketing APIOptional

Live campaign and creative performance; sample mode is available.

GA4 Data APIOptional

Funnel, landing-page and conversion events.

Google Search ConsoleOptional

Organic query and landing-page context.

Microsoft Clarity exportCan be connected

Behaviour evidence and friction signals.

LLM providerRequired

Structured diagnosis, explanation and recommendation drafting.

How output is checked
  • Metric reconciliation across sources
  • Minimum evidence threshold
  • Recommendation confidence
  • Business-impact estimate
  • Duplicate and contradictory recommendation checks
What it will not do on its own
  • Read-only by default
  • No autonomous budget changes
  • Sample and dry-run modes
  • Human approval for exported actions
  • Source-linked recommendations
Metrics it tracks
  • Wasted spending found
  • Customer and lead cost
  • Conversion improvement
  • Tests completed
  • Recommendations accepted
Public-safe examples

What a recruiter
can inspect here.

Real client dashboards and campaign screenshots are not published because of NDA restrictions. These views use fictional or recreated data.

Paid media$3,840

Reallocatable spend

Weak placements and search-term waste identified.

Creative3.7×

Frequency warning

Winning Chili Lime hook is beginning to fatigue.

Performance dashboardRoot-cause treeWaste tableCRO scorecard
Applications

Where this system
was applied.

Public application

Shelizaan

Canadian B2B, retail and multi-location commerce

Paid-media analysis, conversion reviews and marketing reporting.

Public application

Just Kiddin

Indoor playground and family-entertainment company

Campaign analysis, landing-page improvement and test planning.

Public application

Petpooja

Restaurant technology and SaaS

Lead-cost diagnosis, wasted-spend analysis and landing-page CRO.

Public application

Sagar Kathrotiya Show

YouTube and regional-media project

Channel and campaign-performance analysis.

Project names are public. Client screenshots, private data and confidential campaign outputs are not shown because of NDA restrictions.