Problem
Newsletter teams often have performance data spread across exports, campaign reports, audience segments, and spreadsheets. They need to know what changed, which audience is under pressure, and what to do next.
← Case Studies
Marketing intelligence / Data product
Campaign Pulse
Local-first marketing intelligence command center.
Campaign Pulse turns newsletter, campaign, audience, target, and CSV export data into adapter-validated diagnostics, audience pressure signals, target-aware performance reads, and monthly operating reports.
Portfolio-ready / local-first prototype
Next.jsTypeScriptTailwindRechartsLocal JSONlocalStorageFileReader CSVGitHub Actions
Case Snapshot
The strategic brief
The problem, the system response, the available proof, the strategic value, and the intentional boundary.
- 01Problem
- Newsletter, campaign, audience, target, and export data are fragmented across separate operational surfaces.
- 02System
- A local-first command center that validates CSV inputs, normalizes campaign facts, and computes deterministic diagnostics.
- 03Proof
- Live demo, public GitHub repository, focused tests, screenshots, and browser-generated report and export artifacts.
- 04Value
- Turns monthly marketing data into visible decisions about revenue strength, audience pressure, targets, and next actions.
- 05Limitation
- It uses synthetic or browser-local data; it does not connect to a live CRM or email platform.
Solution
Normalize the facts, then make the operating decision visible.
Campaign Pulse turns raw demo newsletter data and uploaded CSV exports into a normalized analytics layer with target-aware diagnostics, audience pressure signals, segment movement intelligence, and monthly operating reports.
What I built
A complete local-first analysis loop.
Adapter contract
One normalized boundary for bundled JSON and flat CSV exports.
Demo JSON adapter
Validates synthetic campaign, newsletter, audience, and target facts.
CSV/export adapter
Merges one-newsletter × one-segment rows into the shared model.
Editable column mapping
Supports inferred aliases, manual overrides, and required-field checks.
Rejected-row diagnostics
Explains invalid rows before a session dataset can be activated.
Client-side CSV upload
Reads local files with FileReader; nothing is sent to a server.
Editable targets
Global, campaign, and segment targets persist in localStorage.
Target comparison chips
On track, watch, and off-track states stay visible in context.
Audience master-detail
All segments remain comparable while one opens into deeper evidence.
Calendar + drawer
Cadence and newsletter detail share the same operating context.
Campaign comparison
Contribution, pressure, strongest sends, and next actions in one read.
Monthly report + export pack
Print-ready memo and browser-generated JSON/CSV artifacts.
CI + deployment prep
Focused tests, GitHub Actions quality gates, and Vercel configuration.
Architecture / System map
Ingestion, intelligence, and presentation stay separate.
Raw source facts do not contain precomputed rates, rankings, diagnoses, or recommendations. Adapters normalize inputs first; deterministic TypeScript utilities compute the intelligence layer after validation.
01Fragmented newsletter and campaign data
02Normalized data model
03Dashboard intelligence
04Monthly decision, report, and export
Inputs
Bundled synthetic JSON or a browser-local CSV export.
Contract
Mapping, validation, rejection reasons, and normalized entities.
Intelligence
Rates, rankings, pressure, movement, targets, risks, and recommendations.
Workspaces
Overview, Calendar, Newsletters, Campaigns, Audience, Insights, Report, and Data.
Key product screens
One dataset, multiple decision surfaces.
The interface is organized as an operating room rather than a generic dashboard. Each workspace answers a different question while preserving the same month, source, target, and detail context.
Data intelligence
Deterministic signals, not AI-generated commentary.
Every read is calculated from normalized local facts. The prototype makes formulas and operating thresholds inspectable, repeatable, and testable.
OR
Unique opens ÷ delivered. A top-level read on subject-line and sender response.
CTR
Unique clicks ÷ delivered. Measures click depth across the delivered audience.
CTOR
Unique clicks ÷ unique opens. Separates post-open content performance from reach.
Conversion rate
Orders ÷ delivered. Connects newsletter delivery to downstream action.
RPR
Revenue ÷ delivered. Compares audience value across sends, campaigns, and segments.
Revenue vs target
Actual revenue is evaluated against editable global or scoped targets.
Pressure / saturation
Send frequency, overlap, and negative signals expose overexposure risk.
Segment movement
Growing, stable, declining, fatigued, and recovering states add direction.
Rejected-row diagnostics
Invalid source rows stay visible instead of silently disappearing.
Target-aware recommendations
Next actions respond to performance and pressure constraints together.
Engineering notes
A credible frontend prototype with explicit boundaries.
The build focuses on product architecture, deterministic analytics, adapter design, and interaction quality without introducing infrastructure the prototype does not need.
- Next.js 14 App Router, React, TypeScript, Tailwind CSS, and Recharts.
- Static local JSON for synthetic demo facts and default target settings.
- Browser FileReader for CSV parsing and session-only dataset activation.
- localStorage for editable target settings only.
- Deterministic analytics utilities with focused adapter, target, upload, and export tests.
- GitHub Actions CI and Git-connected Vercel deployment with no environment variables.
- No backend, database, authentication, API keys, OAuth, or AI/LLM calls.
The implementation workflow also referenced the curated awesome-vibe-coding collection as a development resource. awesome-vibe-coding ↗
Outcome
A portfolio-grade internal tool prototype.
Campaign Pulse demonstrates product thinking, frontend engineering, data modeling, deterministic analytics, adapter design, and UX iteration in one coherent local-first product surface.
Product thinking
Turns fragmented metrics into explicit operating questions and next moves.
Frontend engineering
Coordinates dense analytics workspaces without relying on backend complexity.
Data modeling
Separates source rows, normalized facts, computed intelligence, and presentation.
UX iteration
Balances mission-control scanning with deeper calendar, audience, campaign, and report reads.
