Skip to content

WORKFLOWS.md

Latest commit

 

History

History
50 lines (35 loc) · 4.99 KB

WORKFLOWS.md

File metadata and controls

50 lines (35 loc) · 4.99 KB
Raw

Cross-platform workflows

How the three layers in this repository fit together:

  • Skills are the knowledge: how each platform and each task works. They are grouped by platform (DV360, Google Ads, Amazon DSP, StackAdapt, The Trade Desk) plus shared foundations, operations, measurement, privacy, and reporting.
  • Agents are the roles: planning, trading, optimization, account operations, programmatic ops, ad ops, measurement, analytics, privacy, reporting, client communication, and independent QA. Each agent uses the skills and hands off to the next.
  • Loops are the recurring cycles: bounded routines, run on demand or on a schedule, that monitor and recommend.

The point is one workflow that works across every platform. The agents and loops stay the same. Only the platform skill set changes.

Workflow 1: Launch a campaign

  1. media-planner turns the brief into a plan (objective, KPI, audience, inventory, budget, measurement), using reporting-by-campaign-goal and the platform's targeting and inventory skills.
  2. programmatic-ops-lead opens the launch row, confirms ticket quality, and routes setup, creative, measurement, access, and support blockers.
  3. account-operations-specialist stands up the account and the taxonomy.
  4. programmatic-trader builds the campaign from the plan with the platform's structure, bidding, targeting, deals, and pacing skills.
  5. The pre-launch-qa-gate and floodlight-and-tag-pre-launch-gate loops run with cross-platform-launch-qa: programmatic-ops-lead tracks the go/no-go, account-operations-specialist checks the build, measurement and ad ops verify tags, and qa-scrutinizer independently reviews it. Launch only on clean evidence.

Workflow 2: Run a campaign in flight

  1. The daily-pacing-sweep and anomaly-detection loops watch delivery and the core metrics and surface what needs attention.
  2. The weekly-optimization-pass loop has optimization-specialist diagnose the one binding constraint and propose a single change with its expected impact.
  3. If the recommendation is a large upload or mass edit, bulk-change-safety-gate builds the scope, snapshot, diff, approval, canary, verification, and rollback package.
  4. qa-scrutinizer reviews the recommendation, then the trader approves and applies it. One lever at a time, verified the next cycle.
  5. delivery-and-deal-watch, creative-fatigue-watch, brand-safety-monitor, budget-reallocation, search-term-mining, and audience-performance-review run as the situation calls for them, each ending in a recommendation for approval.

Workflow 3: Run the operations desk

  1. ops-command-center-daily builds the daily desk board across launches, live delivery, tickets, creative, support, access, tracking, incidents, and close risk.
  2. programmatic-ops-lead routes every red or yellow row to the owner and skill: ticket triage, launch QA, platform support escalation, creative review, access governance, measurement, or month-end close.
  3. Support cases, creative rejections, and access changes stay on the board until evidence shows they are resolved or a named outside owner is waiting.
  4. Spend-affecting changes, safety relaxations, access changes, and client-facing updates still require human approval.

Workflow 4: Report and communicate

  1. The weekly-client-report loop has reporting-analyst pull the goal-appropriate report and the so-what, client-communications-lead draft the update, and qa-scrutinizer review it for accuracy and overclaiming before a human sends it.
  2. Ahead of a monthly or quarterly review, the business-review-prep loop assembles the period's results and the path to conversion, reviewed before it reaches the client.
  3. At month end or campaign end, month-end-close reconciles delivery, spend, invoices, discrepancies, make-goods, accruals, taxonomy, and finance handoff before final numbers go out.

Workflow 5: Add or switch platforms

The same agent model and the same loops run on any platform. To move a campaign from one demand-side platform to another, swap the platform skill set the agents draw on:

  • DV360: the dv360-* skills.
  • Google Ads: the google-ads-* skills.
  • Amazon DSP: the amazon-dsp-* and amazon-marketing-cloud skills.
  • StackAdapt: the stackadapt-* skills.
  • The Trade Desk: the ttd-* skills.

The shared foundations, the reporting-by-campaign-goal and path-to-conversion-analysis skills, the agents, and the loops do not change. That is what makes this a multi-platform operating system rather than five separate playbooks.