Case Study

Marketing Operations from intake to attribution —
built as a system, not a process.

A full-stack Marketing Operations system designed for a health technology company scaling from pilot to multi-channel engagement. The brief: replace spreadsheets, Slack threads, and tribal knowledge with a coherent, repeatable campaign engine that an ops team of three can run at scale.

Steps per campaign (intake → closed)

Integrated platforms

Audience lifecycle stages

30→5

Min: human QA time reduction

The Problem

No standard intake — campaigns arrived via Slack, email, or hallway conversation. No structured brief meant Ops spent every Monday reverse-engineering what the requester actually wanted.
No naming convention — UTMs were handmade (spelling errors, missing fields, duplicates), making cross-channel attribution unreliable.
No QA process — campaigns went out without automated checks. Missed suppressions, broken links, and wrong segments were caught by the channel owner after launch — not before.
No feedback loop — no consistent measurement tied campaigns back to business outcomes. The team couldn't answer "which campaigns actually move retention?"

The Approach

Structured intake form in Monday with taxonomy-driven dropdowns. Campaign name auto-generated from selections — no typos, no guesswork.
UTM convention enforced at intake. Every link is tagged consistently for clean attribution across Braze, HubSpot, Short.io, and Lightdash.
Two-layer QA — AI handles the 10 binary checks (did the sync run, does the link resolve, is suppression applied). Human reviews the AI report plus content/compliance in ~5 minutes.
Two-tier measurement — Tier 1 executive scorecard (retention, pipeline, protected revenue). Tier 2 per-campaign reports with anomaly detection.

Key Design Principle

Every campaign follows the same path: structured brief → architecture → build → QA → launch → close. The system makes it easy to do the right thing and hard to skip a step. The UTM convention — not a dashboard — is the attribution layer.

System Architecture

Two tracks. One convention. Clean attribution.

The marketing stack splits into two functional tracks — one for the business-to-business pipeline (employer/partner sales), one for the business-to-consumer engagement layer (member communication). The UTM naming convention is the thread that ties them together.

Two-Track Model

Separate stacks, shared attribution layer

Dimension	B2B Pipeline	B2C Engagement
CRM / Source	Salesforce	Data warehouse
Marketing Automation	HubSpot	Braze
Audience Sync	HubSpot workflows	Hightouch (reverse ETL)
Link Tracking	Short.io	Short.io
Reporting	Lightdash	Lightdash

Naming Convention

The glue that makes attribution possible

{team}_{motion}_{audience}_{quarter}

Team: client / growth / brand
Motion: engagement / retention / acquisition / reactivation
Audience: active-members / at-risk / b2b-clients / prospects
Quarter: auto-generated from date

Example: client_retention_at-risk_2026q3

Standardized Campaign Lifecycle

Every status maps to exactly one lifecycle stage — no ambiguity across teams

Planning

Planned — brief accepted, no work started
Building — assets in production

Execution

In QA — AI checks + human review
Approved — signed off by channel owner
Live — campaign is sending

Closed

Complete — performance read, learnings logged
On Hold — paused by Ops, not dead
Cancelled — documented with reason

Team Coverage Model

Campaigns are owned by one of three marketing teams. Each team's motion maps to specific audience stages:

Team	Primary Motions	Audiences
Client Marketing	Engagement, Retention	B2B Clients, Active Members
Growth Marketing	Acquisition, Reactivation	Prospects, Churned
Brand	Engagement, Awareness	All, broad market

Ownership & Visibility

Single named owner per campaign — not a team, a person
Status tracked: planned → build → QA → approved → live → complete
Weekly ops review: what's live, launching, slipping
Tier 1 dashboard aggregates by motion type — leadership sees health, not volume
Delays flagged proactively — not discovered after launch

Intake & Workflow

Brief to results — no guesswork.

Every campaign follows the same path. Requests come through a structured form. Nothing gets built without a brief. Nothing launches without QA sign-off. Nothing closes without a performance read.

Intake Form

Team fills form in Monday. Campaign name auto-generated from selections.

Triage

Ops reviews: date feasible? Assets ready? Compliance flag? P1/P2/P3 priority assigned.

03 AI

Architecture

AI generates channel plan, audience spec, UTM string, and QA checklist from the brief.

Build

Campaign built in Braze or HubSpot. Audience synced. UTMs logged. Assets staged.

05 AI

AI agent runs automated checks via API. Human reviews the report. Channel owner approves.

Launch

Campaign fires. AI monitors in-flight for anomalies. Day 1/3/7 briefings sent automatically.

07 AI

Monitor & Iterate

AI flags anomalies (open rate drop, unsubscribe spike). Ops reviews Day 7 report. Learnings logged and fed into the next brief.

Live Naming Generator

Select fields to see the campaign name auto-generate — same logic as the intake form.

Requesting Team

Audience Motion

Primary Audience

Generated Campaign Name

client_engagement_active-members_2026q3

Prioritization Framework

P1 — Immediate

High impact + hard-to-reverse. Member-facing sends, partner commitments.

P2 — This Sprint

High impact + reversible. Paid campaigns, internal comms.

P3 — Backlog

Lower impact or long lead time. Scheduled by deadline proximity.

Tools & Stack

Connected tools. One source of truth.

No tool lives in isolation. Every platform passes data to the next. The UTM convention is the thread that ties them all together for clean attribution.

Data Flow

Two parallel tracks — B2B (employer/partner) runs through HubSpot; B2C (member engagement) runs through Braze.

B2B Track — Employer & Partner

Salesforce

CRM
Pipeline data

→

HubSpot

Marketing automation
Campaigns

→

Short.io

Link tracking
UTM attribution

→

Lightdash

Reporting
Pipeline influenced

B2C Track — Member Engagement

Data Warehouse

Source of truth
Member data

→

Hightouch

Reverse ETL
Audience sync

→

Braze

Member engagement
Email, SMS, in-app

→

Short.io

Link tracking
UTM attribution

→

Lightdash

Reporting
Retention + engagement

+ Monday — Intake + campaign tracking + Claude API — AI layer + RRD — Direct mail vendor

Tool Ownership by Role

Tool	Owner	MarOps Role
HubSpot	MarOps	B2B campaigns, build + launch
Data Warehouse	Data & MarTech	Validate outputs
Hightouch	Data & MarTech	Scope + validate syncs
Braze	Shared / MarOps	Member campaigns, QA, launch
Short.io	MarOps	Own all UTM links
Lightdash	Analytics	Scope reports, review
Monday	MarOps	Own the intake queue

APIs Powering Automation

HubSpot API — B2B campaign automation, contact management
Braze REST API — segment validation, subscription status, unsubscribe list
Hightouch API — sync status, timestamp, error count, row validation
Short.io API — link creation, click analytics, UTM validation
Lightdash API — dashboard queries, scorecard updates
Claude API — architecture generation, QA reports, briefings

Audience & Segmentation

Right message. Right person. Right time.

Five lifecycle stages. Every campaign targets one. Overlap is managed by suppression — not audience design. Audiences are built from the data warehouse and synced to the engagement platform via reverse ETL.

Prospect

Motion: Acquisition

Employer/partner decision-makers, not yet a client. Channels: paid, partner email, direct mail, organic.

B2B Client

Motion: Engagement / Retention

Active employer or partner accounts. Channels: email, direct mail, webinar, in-person events.

Active Member

Motion: Engagement

Enrolled members actively using the health platform. Personalized by program milestone and engagement level.

At-Risk Member

Motion: Retention

Disengagement signals: no activity for 7+ days, reduced check-ins, lapsing engagement markers. Trigger-based.

Churned / Inactive

Motion: Reactivation

Members who exited or went fully dark. Lower frequency, higher personalization. Channels: email, direct mail.

Audience Build Process

1.MarOps scopes criteria → submits to Data & MarTech

2.MarTech builds model in data warehouse

3.Reverse ETL syncs to engagement platform on schedule

4.MarOps validates: sync ran, count within tolerance, suppressions applied

5.Segment documented in campaign log before launch

Suppression Logic

Always Applied

Global unsubscribe list (email + SMS)
Subscription group opt-outs
Members in active onboarding sequence
Frequency cap: same motion within [X] days

Campaign-Specific

Audience overlap with concurrent campaigns (P1 wins)
Member-level flags (medical hold, account dispute)

QA System

QA is a system, not a habit.

Habits slip under pressure. Systems don't. Every campaign runs through an AI pre-flight before a human reviews the report — not a blank checklist starting from zero.

Two-Layer Model

AI handles everything binary — did the sync run, does the link resolve, is suppression applied, does the UTM match the convention. The human reviews the AI's report + content/compliance items AI can't assess. Human QA time: 30 min → 5 min.

Layer 1 — AI QA Agent

Automated

🔗

Sync freshness

Last sync <6hrs, zero errors, row count as expected

API

👥

Segment size validation

Count within ±10% of expected — flag if outside

Braze API

🚫

Suppression applied

Unsubscribe list + subscription group status confirmed

Braze API

🔖

UTM present on all links

Every link has source, medium, campaign — no bare URLs

Link scan

✅

UTM convention compliance

Campaign slug matches intake taxonomy exactly

String match

🌐

Short.io links resolve

All shortened links return 200 — no 404s

Short.io API

Layer 2 — Human Review

~5 min

✍️

Subject line + preheader

On-brand, accurate, no typos

Human

📋

Copy + health claims

Factually accurate, compliance reviewed if flagged

Human

🖼️

Imagery + alt text

Renders correctly, alt text present

Human

⏰

Send time + timezone

Correct for target audience geography

Human

👤

Channel owner sign-off

Content approved before launch

Human

Human reviews the AI report, not a blank checklist. Any FAIL from AI stops the launch.

Days 1–30

Human-Only QA

Run the checklist manually. Build baseline error rate. Document current state.

Days 31–60

AI QA Introduced

AI runs automated checks. Human reviews report. Track AI accuracy rate.

Day 60+

Human Review Narrows

AI accuracy verified → human review narrows to content + compliance. Goal: <10 min per campaign.

Measurement Framework

Tier 1 answers "Are we healthy?"
Tier 2 answers "Why?"

Leadership lives in Tier 1 — outcomes, retention trends, protected revenue. Ops and channel owners live in Tier 2 — channel performance, UTM attribution, anomalies.

Tier 1 — Executive Scorecard

Always-on

Metric	Business Value
Member Retention Rate % active at 30/60/90 days	Each retained member ≈ $8K+ annualized savings
At-Risk Reactivation Rate % re-engaged within 14 days	Proxy for churn prevention ROI
B2B Pipeline Influenced Deals touched by marketing	Revenue attribution to marketing
Protected Revenue Retained members × average savings	Estimated value of retention campaigns

Non-revenue campaigns use a proxy: re-engagement event × estimated lifetime value. The $8K figure is based on published industry benchmarks for health engagement programs.

Tier 2 — Campaign Report

Per campaign

Channel	Key Metrics
Email (Braze)	Open rate ±baseline, CTR, unsubscribe, conversion
SMS (Braze)	Delivery, click rate, opt-out rate
Paid Media	CPL, CTR, UTM-attributed pipeline via CRM
Direct Mail	Delivery confirmation, response via short link
In-App	Impression rate, click, action taken post-view

Attribution Chain

CampaignUTM-tagged
Short.io links

→

Short.ioClick recorded
with full UTM

→

CRMConversion captured
with UTM params

→

LightdashAggregated by
utm_campaign filter

→

ScorecardBusiness outcome
attributed to campaign

Because utm_campaign always matches the intake-generated name, the analytics team can query reports with a simple filter — no manual tagging, no cleanup. The convention is the attribution layer.

←→ navigate tabs

Marketing Operations from intake to attribution —built as a system, not a process.

Two tracks. One convention. Clean attribution.

Brief to results — no guesswork.

Connected tools. One source of truth.

Right message. Right person. Right time.

QA is a system, not a habit.

Tier 1 answers "Are we healthy?"Tier 2 answers "Why?"

Marketing Operations from intake to attribution —
built as a system, not a process.

Tier 1 answers "Are we healthy?"
Tier 2 answers "Why?"