Week 5 Progress Report

Node-Based Research Platform

From product framing to a working DAG workflow engine: designing, executing, and measuring multi-channel outreach experiments

Week 4 → Week 5From Strategy to Implementation

Week 4: Product Framing

Defined the target user, value proposition, and integration scope for an operations-connected A/B platform.

"We know who it's for and what it should do. Now we need to build it."

→

Week 5: Working System

Implemented a node-based DAG workflow editor with a backend execution engine that runs multi-step outreach campaigns.

"Practitioners can now visually design, execute, and monitor experiment workflows in real time."

The Leap: Week 5 transforms the conceptual platform into a functional system. The DAG engine allows campaigns to branch, pause on events, evaluate conditions, and track per-node metrics — making A/B experiments executable, not just designable.

Week 4 BaselineStarting Point for This Week

Target user defined: marketing managers at large appointment-based businesses

Product positioned as an ops-integrated A/B research platform

Core module boundaries defined: Strategy, Experiment, Analytics

Campaign CRUD, participant allocation, and outreach execution already functional

Missing: a visual workflow authoring tool and an engine to run multi-step strategies as DAGs

Week 5 ProgressThree Implementation Pillars

1. Visual DAG Workflow Editor

Built a three-column campaign editor using React Flow: a node palette on the left, a drag-and-drop canvas in the center, and a per-node configuration panel on the right. Practitioners can visually compose multi-step outreach strategies.

2. DAG Execution Engine

Implemented a backend DAG executor that traverses the workflow graph, executing each node in sequence. Supports branching via condition checks, pausing via event gates and wait delays, and A/B splitting via allocation nodes.

3. Visual Audience Targeting

Created a criteria-based audience filter with AND/OR condition groups, behavioral targeting (visit count, spending, appointment status), and live audience preview with matched guest counts.

Node Type Registry10 Specialized Workflow Nodes

👥

Customer Selector

Audience targeting entry point

✂

Split Group

A/B allocation by percentage

✉

AI Email

AI-generated email outreach

💬

AI SMS

AI-generated text message

📞

AI Voice

AI-generated voice call

❓

Condition Check

Branch on guest behavior

⏳

Wait Delay

Pause for N hours/days

🔔

Event Gate

Pause until external event

📋

Manual Task

Assign to human operator

✅

Record Conversion

Mark experiment outcome

CapabilitiesWhat the Platform Can Do Now

DDAG Workflow Design

Drag-and-drop canvas with React Flow. Connect nodes with edges to compose branching, multi-step outreach strategies visually.

EAutomated Execution

Backend engine traverses the graph, executing each node. Participants flow through the DAG with state tracking at every step.

MReal-Time Metrics

Per-node execution counters (entered, exited, success, failed) collected atomically and polled by the frontend for live dashboard updates.

AAudience Segmentation

Criteria-based targeting with AND/OR groups, behavioral conditions, and live preview showing matched guest count and sample profiles.

ComparisonWeek 4 vs Week 5

Area	Week 4	Week 5
Workflow Authoring	Conceptual module boundary	React Flow DAG editor Built
Execution Engine	Not implemented	DAG executor with 10 node types Built
Node Types	Channel concept only	10 typed nodes with config forms Built
Live Monitoring	Not available	Per-node metrics with polling Built
Audience Targeting	Tag-based criteria	Visual filter with AND/OR groups Enhanced
Campaign Lifecycle	Basic CRUD	Full state machine with event listeners Enhanced

ArchitectureResearch Platform System View

Editor

Node Palette

Workflow Canvas

Config Panel

↓

Engine

DAG Executor

Event Listener

Node Metrics

↓

Channels

AI Voice

AI SMS

AI Email

Manual Task

↓

Analytics

Conversion Funnel

Group Comparison

Statistical Significance

↓

Data

PostgreSQL (Multi-tenant)

Workflow Graph (JSONB)

Event Listeners

RoadmapWhat Comes Next

Week 3

A/B platform spec and interactive demo

Week 4

ICP refinement and product framing

Week 5

DAG workflow editor + execution engine + audience targeting (current)

Week 6

End-to-end pilot: run a real campaign through the DAG pipeline

Week 7

Analytics validation and conversion attribution refinement

Final

Stakeholder review with evidence from pilot experiments

Week 5 Complete

← Back to Timeline

The research platform now supports the full cycle: design workflows visually, execute them as DAGs, and monitor per-node metrics in real time. Week 6 will focus on running an end-to-end pilot campaign.