Outcome
By the end of this lesson, you will be able to use the savings overlay to identify reclaimable spend in the Sankey, deep-link to filtered recommendations, and combine the savings overlay with the unattributed-teams overlay for diagnostic depth.
| Tier | Architect |
| JTBD | ”Find the money I can recover this week directly from the cost flow, without juggling separate recommendation pages.” |
| Personas | FinOps Lead · Platform Engineer · Engineering Leader |
| Prerequisites | M3.8.L1-L3 (Trend vs Flow, layouts, drill) · T1: Recommendations |
| Time | 9 minutes |
| Bloom verb | Use (Apply), Deep-link (Apply), Combine (Analyze) |
1. Concept
Two things can be laid over the diagram.
The first marks the flows that have money waiting to be reclaimed, meaning open recommendations. The second marks resources with no team against them.
Between them they change the question the picture answers, from where the money went to where you should go next.
WITHOUT OVERLAYS: Sankey shows flow size only Each flow's color = neutral User sees composition; must navigate elsewhere to act
WITH SAVINGS OVERLAY: Flows with open recommendations get a red stripe Hover any flow: inspector shows "$X reclaimable" if any Click the callout: deep-link to filtered Recommendations
WITH UNATTRIBUTED OVERLAY: Flows containing untagged resources are highlighted yellow Distinguishes "where the money goes" from "where the money is unattributed"
BOTH OVERLAYS ON: Red-and-yellow striped flows = double signal (untagged + reclaimable)What “reclaimable” means
RECLAIMABLE = sum of potential savings from all OPEN recommendations on resources in this flow
For a flow showing "prod-aws-us-east-1 → EKS → team-platform": Reclaimable = SUM(recommendations.estimated_savings_usd) FROM open_recommendations WHERE resource IS IN scope(this flow)The number on the callout is the dollar amount of potential savings that have not yet been applied. Applying the recommendation would convert reclaimable → realized.
Filtering recommendations from overlay
The savings overlay’s most useful feature: deep-linking to filtered recommendations.
USER CLICKS the "$2,400 reclaimable" callout on a flow
OPENS Recommendations page with filters pre-applied: cloud_account_id = prod-aws-us-east-1 resource_type = EKS team = platform (if applicable in the drill path) status = open
RESULT: Recommendations list shows only the 8 specific recs that contribute to the $2,400 number. User triages directly.
NO MANUAL FILTER CONSTRUCTION required.This is the bridge between “Cost Flow analysis” (where the money goes) and “Recommendations action” (what to do about it). Without the deep-link, the user would have to read the flow, open Recommendations separately, construct matching filters, and confirm they match. The deep-link compresses 30 seconds of navigation into one click.
Limitations
SUPPORTED DEEP-LINK FILTERS: resource_type cloud_account_id team_id (limited; via team scoping)
NOT YET SUPPORTED: provider (callout shows; deep-link hidden) region (callout shows; deep-link hidden) service_name (callout shows; deep-link hidden) purchase_type (callout shows; deep-link hidden) resource_uid (drill into resource detail page directly)
WHEN A CALLOUT IS HIDDEN: The savings exist but cannot be directly filtered to Workaround: open Recommendations and filter manuallyThis is a known limitation; future versions are likely to support more dimensions. The current set covers the most-common drill paths.
Unattributed teams overlay
SEPARATE OVERLAY: Unattributed teams highlight
When toggled on: Resources without a team tag are highlighted in yellow Flows that include any untagged resources show the yellow tint Hover: inspector shows the count and total cost of untagged resources in the flow
USE CASES: - Tag-coverage audits - Identifying which dimensions have the most untagged spend - Finding orphan resources (often correlated with low team-tagging)The unattributed overlay is the visual companion to the Tag Coverage widget (M3.5.L4). Where the widget gives a percentage, the overlay shows where the percentage gap lives geographically in the Sankey.
Combining overlays
BOTH OVERLAYS ON SIMULTANEOUSLY:
Red-striped: open recommendations (reclaimable savings) Yellow-highlighted: untagged resources Red-and-yellow: both (recommendations + untagged)
INTERPRETATION: Red only: known waste; act via recommendations Yellow only: tagging gap; address via tagging discipline Red and yellow: double opportunity: untagged AND optimization available. Often orphan resources from departed users; tag + apply rec or just delete.
The red-and-yellow flows are usually the highest-leverage.When to use the savings overlay
SCENARIO USE OVERLAY?──────────────────────────────────────────────────────────────────Weekly Operate review Yes: find this week's action items
Monthly cost review for leadership No: overlay adds noise to executive view; show the number separately
Quarterly optimization sprint Yes: primary tool for finding savings
Anomaly investigation Maybe: sometimes the anomaly is from open recs not yet applied
Initial onboarding / exploration Yes: surfaces value immediatelyThe overlay adds visual complexity. Use it when the additional information helps; turn it off when a clean view is needed.
Example overlay state
COST FLOW with savings overlay, BY SERVICE layout:
Provider → Account → Service → Team
AWS → prod-us → EKS → team-ml-infra ████░░░░ ███████░ ████░░░ ▒▒▒▒▒▒▒▒ ▒▒▒▒▒▒▒▒ ▒▒▒▒▒▒▒ (red stripes: $4.2K)
AWS → prod-us → RDS → team-platform ████░░░░ ███░░░░░ ██░░░░░ (red stripes: $1.1K)
GCP → dev → Compute → (untagged) ██░░░░░░ █░░░░░░░ █░░░░░░░ ░░░░░░░░ ░░░░░░░░ ░░░░░░░░ (yellow: $480) (red+yellow: $480 from untagged compute, also has open recs)
LEGEND: ▒▒▒ = red stripe (reclaimable savings) ░░░ = yellow tint (untagged resources)The visual encoding makes opportunities pop out of the Sankey.
Weekly Operate ritual
The savings overlay is at the center of a productive Weekly Operate ritual:
WEEKLY (20 minutes): T+0 Open Cost Flow, default layout, this-week scope T+1 Toggle savings overlay T+2 Identify the largest red-striped flow T+3 Click the callout → filtered Recommendations T+5 Triage 3-5 recommendations (apply / dismiss / investigate further) T+10 Repeat for next red-striped flow T+18 Document acted recommendations in team wiki T+20 Close loop
OUTCOME over time: Reclaimable savings decline as recs are applied Tag coverage improves as untagged resources are addressed The overlay becomes "the thing I scan weekly"How ZopNight uses the overlays
The overlay states are URL-encoded alongside layout and drill. The toggle states persist across Sankey reloads. The reclaimable-savings calculation joins the Sankey query result with open_recommendations; the untagged-resources calculation cross-references the team-tag inventory.
For performance, the overlay queries are cached for a few minutes; recommendations don’t change minute-to-minute, so a slight delay is acceptable in exchange for fast renders.
2. Demo
A weekly Operate cycle using the overlay:
TIME: Monday morning, 9:15 AMGOAL: This week's recommendations to triage
T+0 Open Reports → Cost FlowT+5 s Toggle "Savings overlay" → onT+10 s Scan the Sankey for red-striped flows
Three biggest: 1. AWS → prod-us → EKS → team-platform $2,400 2. AWS → prod-us → RDS → team-product $1,100 3. GCP → prod → BigQuery → team-data $750
T+30 s Click the EKS callout ($2,400)T+35 s Recommendations page opens filtered: cloud_account_id=prod-aws-us-east-1, resource_type=EKS Sees 8 open recommendations totaling $2,400/mo
T+1 min Triage: - 5 right-sizing recs (low risk) → apply now - 2 idle-cluster recs (need owner approval) → assign - 1 spot-instance switch (medium risk) → defer
T+3 min Apply the 5 right-sizing recs (one-click each)
T+5 min Switch back to Cost Flow; click the RDS callout ($1,100)T+5:30 Recommendations opens for RDS in prod-us Triage: 4 recs; 3 apply now, 1 defer
T+8 min Click the BigQuery callout ($750)T+8:30 BigQuery recs (snapshot lifecycle, query optimization) Triage: 2 recs; both deferred (need data team owner)
T+10 min Switch on Unattributed overlay; notice yellow on flow: GCP → dev → Compute → (untagged) 12 compute resources without team tags
T+11 min Click into the untagged section Identify the resources Assign team tags or flag for cleanup
T+15 min Document in team wiki: "This week: applied 8 recs ($2.4K + $1.1K = $3.5K/mo recoverable); tagged 12 GCP compute resources."
ELAPSED: 15 minutes for $3,500/mo of recovered savings + tagginghygiene.3. Hands-on (5 min)
Use the savings overlay:
OPEN Reports → Cost Flow → toggle Savings overlay
LARGEST RED-STRIPED FLOWS (top 3 you see): 1. __________ Reclaimable: $__________ 2. __________ Reclaimable: $__________ 3. __________ Reclaimable: $__________
TOTAL VISIBLE RECLAIMABLE: $__________
PICK ONE FLOW; click its callout: Recommendations opened? Yes / No Filter scope: __________ Number of recommendations: _____
ACTION you can take this week: □ Apply N recommendations (low-risk) □ Assign owner for high-risk recs □ Defer specific recs (note why) □ Combine with team meeting
UNATTRIBUTED OVERLAY: Toggle on. Largest untagged flow: __________ Untagged amount: $__________ Tagging plan: __________If you have no red-striped flows, your team is in optimal state: recommendations are kept current. Time to switch focus to tagging hygiene or unit-economics improvements.
4. Knowledge check
Q1
A red-striped flow indicates:
A. A cloud provider billing error somewhere in the flow
B. Open recommendations exist on resources in this flow
C. A particularly high-cost flow within the cascade
D. An anomaly that the detector flagged on that flow
Show answer
Correct: B. The “$X reclaimable” callout deep-links to the matching recommendations. The stripe is the visual cue; the click is the action. Red stripe = reclaimable savings via open recommendations. The visual cue is what makes the overlay productive.
Q2
The deep-link from a savings callout:
A. Opens Recommendations filtered to match the flow’s drill path (cloud_account_id, resource_type, etc.)
B. Opens the Recommendations page entirely unfiltered, so you have to re-filter it yourself again
C. Just shows the count of open recommendations without linking anywhere at all that is useful
D. Opens the resource list filtered to that flow, but never the recommendations themselves
Show answer
Correct: A. Saves the user from manually constructing the filter. The filter scope matches the callout’s reclaimable amount. Filtered to match. The bridge between Cost Flow and Recommendations.
Q3
The deep-link works for which dimensions?
A. Every single dimension available in the cascade
B. None of them at all; the linking is manual only
C. Only the team and the cloud account dimensions
D. resource_type and cloud_account_id are well-supported
Show answer
Correct: D. Other dimensions (provider, region, service, purchase_type) show the callout but the deep-link is hidden; user opens Recommendations and filters manually. Future versions are likely to expand the supported set. Limited dimension support is the current state. The most-common dimensions are covered.
5. Apply
Toggle the savings overlay in Reports → Cost Flow. The toggle is in the upper-right of the Sankey view. The unattributed overlay is in the same toggle group.
For weekly Operate cadence, build the savings-overlay scan into your routine. 15-20 minutes per week is sufficient to capture most recoverable savings.
Related lessons
- L1: Trend vs Flow
- L2: Five layout dimensions
- L3: Drill, breadcrumb, back-out
- T1: Recommendations module (Operator tier)
- T3.M3.5.L4: Tag coverage
Glossary terms touched
Savings overlay · Reclaimable · Unattributed overlay · Deep-link · Open recommendation
Module quiz
Complete M3.8 → 10-question module quiz. Track 3 complete.