Campaign Intelligence Dashboard
A triage-first dashboard tab that turns raw campaign data into health signals, ranked recommendations, and a cross-tab filtered workspace.
Context
Media buyers managing dozens of live campaigns don't have time to open every campaign and manually check delivery pace, budget consumption, and creative performance. The Intouch Ant dashboard added a Campaign Intelligence (CI) tab designed to answer one question first: "What needs my attention right now?"
The Problem
The existing dashboard surfaced raw activity, but left prioritization entirely to the user — there was no signal distinguishing a campaign that's healthy and on-track from one that's about to miss its delivery target or burn through budget early. Users had to build their own mental model of "normal" by cross-referencing numbers across multiple campaigns.
Design Goals
- —Give users an at-a-glance triage view: how many campaigns are healthy, how many need attention, how many are wrapping up soon, how many are waiting on approval.
- —Surface per-campaign health and forecast signals directly in the list, so users don't have to open each campaign to know if it's on track.
- —Go one step further than "here's the data" — provide a recommendations rail that translates signals into specific suggested actions, ranked by urgency.
- —Connect the dashboard to the rest of the app: selecting campaigns here should carry that context into the Ad Groups and Ads views.
Key Design Decisions
Stat-card summary row
Four cards across the top — Active Campaigns, At Risk, Completing This Week, Pending Approval — give an immediate read on the health of the whole portfolio before any scrolling. This frames the rest of the tab: "X campaigns need a decision from you today."
Health & Forecast columns in the campaign table
Rather than a separate "alerts" panel disconnected from the data, health and forecast indicators were added as columns in the same table users already scan for names, delivery status, and progress. Health and forecast are derived signals (not raw fields) — computed from the underlying campaign data — so the table communicates interpretation, not just numbers.
The table reuses the same configurable PortalTableComponent from Sponsored
Campaigns (Saved Views, density toggle, sticky columns), with a CI-specific column set
seeded as its default view — so the dashboard gets a tailored layout without
maintaining a separate table implementation.
Recommendations rail
A dedicated right-hand rail surfaces specific, ranked suggestions — categorized as urgent, optimize, or explore — each with a confidence indicator and Dismiss / Act controls. This is the layer above "here's a chart": it's the product taking a position on what the user should do next, while still leaving the final call to them (dismissible, not automatic).
Cross-tab selection
Selecting one or more campaigns updates the Ad Groups and Ads tabs to show only items belonging to those campaigns, with the tab label and a removable selection badge reflecting the active filter. This turns "I'm worried about these three campaigns" into a filtered workspace across the whole app, rather than a one-off view.
A Technical Note Worth Including
Early implementation hit a subtle but product-relevant bug: a layout grid configuration caused an infinite re-render loop under certain conditions, which would have manifested to users as a frozen or sluggish dashboard. Diagnosing and fixing this is a good example of design intent meeting implementation reality — the fix required no visual change, but was essential to the feature being usable at all.
Outcome
A dashboard tab that goes from "data about campaigns" to "a prioritized worklist," with a direct path from "what's wrong" (stat cards / health column) → "what should I do about it" (recommendations rail) → "let me focus on just these" (cross-tab filtering).
What This Demonstrates
- —Designing information hierarchy for triage, not just display.
- —Turning derived/interpreted signals (health, forecast) into first-class UI elements alongside raw data.
- —Designing a recommendation surface that's actionable but not prescriptive (dismissible suggestions, confidence levels).
- —Cross-view interaction design — selection state that has consequences beyond the component it lives in.
- —Comfort working close enough to implementation to catch issues that affect real-world usability, not just visual design.