Recommendations
Guard generates AI-powered recommendations to help you reduce costs, eliminate waste, and improve your cloud posture. Each recommendation includes an estimated impact so you can prioritise the highest-value actions.
Recommendation categories
| Category | What it recommends |
|---|---|
| Rightsizing | Downsize over-provisioned instances based on actual utilisation data |
| Scheduling | Add Schedules to resources running outside business hours without automation |
| Commitment | Purchase or modify Reserved Instances and Savings Plans based on usage patterns |
| Waste elimination | Delete or decommission unused resources (idle instances, unattached volumes, unused IPs) |
How recommendations are generated
frugally.app analyses:
- Resource utilisation — CPU, memory, and network metrics from CloudWatch
- Usage patterns — When resources are active vs idle
- Cost data — Current spend and historical trends from Intelligence
- Scan findings — Issues discovered during Guard Scans
- Commitment coverage — Gaps in RI/Savings Plan coverage from Commitment Utilisation
Recommendations are refreshed daily and reflect the most recent data available.
Viewing recommendations
The recommendations page shows a prioritised list:
| Column | Description |
|---|---|
| Category | Rightsizing, Scheduling, Commitment, or Waste |
| Description | What the recommendation suggests |
| Affected resources | The resources involved |
| Estimated monthly savings | How much you could save by acting on this recommendation |
| Confidence | How confident frugally.app is in the recommendation (High, Medium, Low) |
| Status | New, Accepted, Dismissed, Snoozed |
[SCREENSHOT: recommendations-list.png -- Recommendations list sorted by estimated savings]
Acting on recommendations
Accept
Accepting a recommendation signals your intent to act on it. Depending on the category:
| Category | What happens on accept |
|---|---|
| Rightsizing | Creates a draft Guard Project with the proposed instance type change |
| Scheduling | Opens the Schedule creation form pre-filled with the suggested configuration |
| Commitment | Provides a summary you can use when purchasing RIs/Savings Plans in the AWS console |
| Waste | Creates a draft Guard Project proposing decommission of the resource |
Accepted recommendations are tracked in the Cost Posture optimisation dimension.
Dismiss
Dismissing a recommendation removes it from the active list. You must provide a reason:
| Reason | Example |
|---|---|
| Not applicable | "This resource is intentionally over-provisioned for burst capacity" |
| Already planned | "We are migrating this workload next month" |
| Risk too high | "Production database — cannot risk downtime for rightsizing" |
Dismissed recommendations do not count against your Cost Posture score.
Snooze
Snoozing hides a recommendation for a configurable period (7, 14, 30, or 90 days). After the snooze expires, the recommendation reappears if it is still relevant.
Estimated impact
Each recommendation shows an estimated monthly savings figure. This is calculated based on:
- Current cost — What the resource costs now
- Proposed cost — What it would cost after the recommendation is applied
- Savings — The difference, projected monthly
Estimated savings are projections based on current pricing and usage. Actual savings may vary based on usage changes, pricing updates, and implementation timing.
Best practices
- Review weekly — Check recommendations as part of your regular FinOps routine
- Prioritise by impact — Sort by estimated savings and tackle the top items first
- Validate before acting — For rightsizing, review CloudWatch metrics to confirm low utilisation is consistent, not a temporary lull
- Dismiss with reasons — This helps your team understand why a recommendation was not acted on and prevents it from being re-raised
- Track accepted recommendations — Monitor your Cost Posture to see the cumulative impact of implemented recommendations