Approvals
Approvals (previously called Actions) is Vendo’s human-in-the-loop layer for AI recommendations. The system generates findings — KPI summaries, trend analysis, anomalies, budget shifts, audience changes, pipeline health alerts — and queues them for your review before anything executes.
Open Approvals from the sidebar.
How it works
- Vendo monitors your connected data on a configured schedule.
- When something noteworthy is detected, a recommendation appears in the pending queue.
- You review the recommendation, its reasoning, and supporting data.
- Approve to trigger the recommended action, or Dismiss to skip it.
This keeps AI useful without black-box automation — especially important for agencies and enterprise teams who need explainability and accountability.
Configure Approvals
From the Approvals settings panel you can configure:
| Setting | Options |
|---|---|
| Cadence | Daily, weekly, monthly, on-demand, or anomaly-triggered |
| Domains | Calculated properties, transformations, segments, pipelines, cross-source, activation |
Each domain scopes which types of recommendations Vendo generates.
Recommendation types
| Type | Example |
|---|---|
| KPI summary | ”ROAS dropped 18% on Meta this week” |
| Trend analysis | ”New customer acquisition trending down for 3 consecutive weeks” |
| Anomaly | ”Shopify order volume 2.4σ below 30-day average” |
| Recommendation | ”Shift $2,400 from Google Display to Meta retargeting” |
| Audience change | ”High-LTV segment grew 12% — consider expanding lookalike” |
| Pipeline health | ”Stripe sync failed twice in 24 hours” |
| Transformation performance | ”LTV model accuracy degraded — retrain recommended” |
Approve vs dismiss
Approve tells Vendo to execute the recommended action — this may create a Task, trigger a sync, adjust a configuration, or queue a downstream workflow depending on the recommendation type.
Dismiss removes the recommendation from the queue without action. Dismissed items are logged for audit purposes.
Related
- Automations — operationalize approved recommendations
- Agent — ask questions and generate ad-hoc analysis
- Memory — track what changed after an approval
- Reports — scheduled AI commentary