Building an Internal Campaign Operations System

The challenge wasn't sending email. The platform already had send infrastructure. The problem was everything that happens before the send: turning a pile of contacts into a clean, reviewable, reusable audience that a non-technical user could trust.

Audience prep was fragmented. Marketing teams pulled CSV exports, filtered them in spreadsheets, and re-imported the cleaned list. Tags were inconsistent — the same concept was tagged three different ways in three different campaigns. Segments existed but weren't surfaced clearly enough for marketing users to build them confidently.

Campaign state was tracked with scattered boolean flags — is_draft, is_sent, is_scheduled — which meant the UI had to guess what action to show next. Drafts that were actually mid-send got the wrong actions. Failed campaigns looked the same as completed ones until you opened the report.

Reporting visibility was thin. Teams could see whether a campaign sent, but not the audience size at send time, what came back, or which records failed. Without that loop, every campaign felt like it ran in the dark.

The goal was to fix the prep layer, not rebuild the send layer.

I built the campaign workflow around three concepts kept deliberately separate: lists (hand-picked groups), tags (loose labels), and segments (rule-based audiences). Marketing users already thought of them differently — collapsing them into one "audience filter" abstraction would have made the data model cleaner but the UX worse.

On top of that, I modelled campaign state as an explicit state machine — draft → scheduled → sending → sent → completed, with failed as a side branch from sending. Each state had its own allowed actions, its own badge, and its own report view. Replacing the boolean-flag soup with one status column collapsed a lot of conditional logic.

The audience builder pulled from any combination of lists, tags, segments, manual selections, and CSV imports — deduplicated by contact ID, with a visible count that updated as you added or removed sources. Lists were promoted to first-class reusable units: create once, reuse across campaigns, edit in place.

On the backend, campaign and audience APIs returned predictable shapes: counts at the top, contact previews paginated below, status and timestamps explicit. Angular consumed those endpoints into screens designed for marketing users — readable status badges, audience-size feedback while building, and confirmation steps on anything irreversible.

CSV import got its own validation layer: required fields checked, duplicates flagged before save, and a preview step before contacts hit the database. Reporting hooked into the send pipeline directly, capturing audience size, send activity, and engagement at the campaign level.