Command Center

Command Center is the recommendation workbench for review, approval, rejection, and queue execution. The guided workflow uses 5 phases with 24 lanes.

Prerequisites

• Connected Amazon account
• Selected marketplace
• Contributor or Admin role for approve/reject actions

Expected outcome: Command Center actions and workflow controls are available.

Quick Start

1. Open Command Center.
2. Start in Triage and process critical lanes first.
3. Move through Defend, Attack, Optimize, then Maintain.
4. In each phase, open a lane, review the workbench details, then approve, or skip.
5. Verify applied outcomes in Activity Log.

Expected outcome: Recommendations are processed in the intended risk-first sequence.

Detailed Workflows

Workflow: Run the Guided Workflow Correctly

1. Process phases in order: Triage -> Defend -> Attack -> Optimize -> Maintain.
2. Within a phase, process non-zero lanes first.
3. When a phase is complete, continue to the next unlocked phase.
4. If campaign filters are active, check unfiltered counts before declaring a phase complete.
5. Repeat until all phases show zero global pending items.

Expected outcome: High-risk issues are resolved before growth and maintenance work.

Workflow: Process One Lane End-to-End

1. Select a phase.
2. Select a lane in that phase.
3. Review recommendation details, evidence panels, and the proposed mutation.
4. Approve when valid, or skip/reject when it violates policy.
5. Watch for queued status and continue until lane count reaches zero.
6. Validate resulting state in Activity Log.

Expected outcome: Lane backlog is cleared with consistent operator decisions.

Workflow Taxonomy (Current Model)

• Workflow phases: 5
• Active workflow lanes: 24
• Recommendation categories supported by the engine: 23
• Recommendation categories shown as guided lanes: 21

Expected outcome: You can distinguish guided workflow lanes from total supported recommendation categories.

Workflow: Review a Recommendation in the Workbench

1. Open the target lane and select a recommendation.
2. Confirm the entity identifiers shown in the workspace header.
3. Review evidence, impact estimates, and any strategy or performance panels shown for that type.
4. Use explain-on-demand details when you need more reasoning before acting.
5. Approve, reject, or move to the next item.

Expected outcome: Each decision is made from the current split-view workbench rather than a lightweight card-only review.

24 Guided Actions (Triggers, Cadence, Outcomes)

Cadence roles:

• Sentinel (hourly) - risk and health monitoring
• Operator (daily) - execution-heavy optimization
• Strategist (weekly) - structural and scale planning

Phase	Guided action	What triggers it	Cadence	What you get when accepted
Triage	Critical Alerts	Severe risk signals like ACOS spikes, budget exhaustion, or urgent guardrail breaches	Sentinel (hourly)	Immediate protective fix is queued so you can prevent revenue loss quickly
Triage	Legacy Protection	Mature/legacy campaigns that should not be changed aggressively	Strategist (weekly)	Campaign is put into protected mode so risky optimization recommendations are no longer generated for it
Defend	Downbids	High-spend keywords or targets with poor efficiency	Operator (daily)	Bids are lowered to reduce waste and protect margin
Defend	Negative Keywords	Search terms spending with weak or zero conversion	Operator (daily)	Blocking negatives are added so low-quality traffic stops triggering ads
Defend	Product Page Shield	Competitor pressure on your product detail pages	Operator (daily)	Defensive targeting is created to help hold your PDP traffic
Attack	Upbids	Proven winners that are under-scaled	Operator (daily)	Bids are increased to capture more profitable volume
Attack	Harvest	Converting search terms found in broader/discovery traffic	Operator (daily)	High-performing terms are promoted into tighter targeting and cleaned up at source
Attack	ASIN Harvest	Strong product-level opportunities against specific ASINs	Strategist (weekly)	New product targets are created to capture those opportunities directly
Attack	Category Discovery	Early signs of conversion in untapped categories	Strategist (weekly)	New category-level targeting is launched to test expansion areas
Attack	Category Scaling	Existing categories already showing good return	Strategist (weekly)	Category targeting is expanded to scale proven pockets of demand
Attack	Scale Budgets	Profitable campaigns hitting budget ceilings	Operator (daily)	Daily budgets are increased so winning campaigns do not cap early
Attack	Anchor Keywords	Keywords with outsized contribution to revenue efficiency	Strategist (weekly)	Priority “anchor” terms are identified for focused investment
Attack	Focus Campaigns	Anchor terms that need tighter control and isolation	Strategist (weekly)	Focused campaign structures are created for better control of bids and placements
Attack	Auto Campaign Creation	Gaps in discovery coverage	Strategist (weekly)	New discovery campaigns are launched to find additional winners
Optimize	Dayparting	Clear time-of-day/day-of-week performance differences	Operator (daily)	Schedules are applied so spend shifts toward high-performing windows
Optimize	Placement Optimization	Imbalance across top-of-search, product page, or rest-of-search performance	Operator (daily)	Placement multipliers are adjusted to improve return by placement
Optimize	Variation Optimization	Meaningful performance gaps between creative/ad variations	Operator (daily)	Underperforming variations are reduced so delivery concentrates on stronger creatives
Optimize	Cut Budgets	Campaigns consistently underperforming vs target	Operator (daily)	Budgets are reduced to limit waste and reallocate spend
Optimize	Budget Rules	Repeating budget patterns that can be automated	Operator (daily)	Budget automation rules are created for hands-off control
Maintain	Naming	Campaign naming drift from your standards	Strategist (weekly)	Names are standardized to improve reporting and team clarity
Maintain	Ad Group Naming	Ad-group naming drift from your standards	Strategist (weekly)	Ad-group names are standardized for cleaner structure and analysis
Maintain	Allocation	Portfolio mix drifting from intended discovery/scale balance	Strategist (weekly)	Clear allocation guidance is recorded so spend mix is corrected intentionally
Maintain	Structure	Structural issues that reduce control or quality	Strategist (weekly)	Structural improvements are applied to keep the account clean and manageable
Maintain	Health Alerts	Non-urgent warnings/info signals (for example cannibalization or hygiene findings)	Sentinel (hourly)	Preventive fixes are queued before issues become critical

Expected outcome: Every guided action has a clear trigger, role cadence, and acceptance behavior.

Recommendation Categories Not Exposed as Standalone Lanes

• Product Split recommendations run through guardrail/product-split flows, not as a dedicated lane.
• Campaign Classification recommendations update campaign role metadata, not as a dedicated lane.

Expected outcome: Operators do not assume every supported recommendation category appears as a lane.

Common Errors

Error: Phase looks complete but work remains

1. Clear campaign/type filters.
2. Compare filtered and unfiltered phase counts.
3. Process remaining global items.

Expected outcome: Completion state reflects actual backlog.

Error: Bid or budget recommendations appear in the wrong lane

1. Verify whether the recommendation intent is protect (Defend) or scale (Attack) for bids.
2. Verify whether the budget action is increase (Attack) or decrease (Optimize).
3. Re-open the correct lane and continue processing.

Expected outcome: Recommendations are handled in the correct strategic context.

Error: Approve action does not show immediate result

1. Confirm status transition (pending_approval -> queued -> final state).
2. Open Activity Log for execution details.
3. If failed, resolve root cause and retry from Command Center.

Expected outcome: Execution status is understood and acted on correctly.

FAQ

How should I move through Command Center?

The active model is phase/lane navigation: 5 phases and 24 lanes.

Why do some recommendation types appear in more than one lane?

Some recommendation categories appear in more than one lane based on intent (for example defend vs attack bids, increase vs decrease budgets, critical vs non-critical alerts).

What happens after I approve something?

Approval generally queues execution; final applied outcome is confirmed in Activity Log.

Last Updated

• Last updated: 2026-03-18
• Version assumptions: current workflow config, within-phase priority order, Command Center workbench shell, recommendation decision workspace, and recommendation handler registry