After 12 months of dashboard reviews with FinOps teams, the same 8 patterns separate dashboards that drive cost decisions from those that just look pretty. Here's the blueprint.
Lakshmi Kiranmai Guduru
Founder, CARTIEAI
You open the page. You see this month's spend, last month's spend, a pie chart by service. You stare at it.
You close the tab.
Nothing happened. No decision was made. No money was saved.
After reviewing 60+ FinOps dashboards over the past 12 months — Datadog screens, Looker boards, custom Grafana setups, Snowflake's own usage UI, internal Tableau decks — I started seeing the same 8 patterns separate dashboards that drive decisions from dashboards that just exist.
Here they are. Steal them.
The mistake: Showing today's total spend.
The fix: Show how many dollars are anomalous — i.e. how much of today's spend deviates from your normal.
A $48K spend day is meaningless. A $48K day with $11K in anomalies is an emergency.
A simple lane:
Today: $48,210 · Normal: $37,400 · Anomalous: $10,810 (4 incidents)
Click any incident → drills into the service, the resource, and the SQL or workload that triggered it.
Decision triggered: "Page the on-call. Stop the bleed."
The mistake: Showing total Snowflake spend.
The fix: Showing cost per active customer or cost per 1M API requests.
Total spend grows with your business. Unit cost reveals whether your business model is healthy. If your Snowflake spend grew 40% but customers grew 50%, your unit cost is down — that's good. Most dashboards hide this.
A simple tile:
Cost / Active Customer (this month)
$2.31 ▼ -8% MoM
Decision triggered: "Our infra is getting more efficient — we can lower pricing or reinvest."
The mistake: A single forecast line.
The fix: A forecast line plus the 80% confidence interval shaded behind it.
A single line implies precision your model doesn't have. A confidence band is honest — and operationally useful: the upper bound is your guardrail trigger.
If actual spend pierces the upper band, that's the moment to alert.
Forecast (P50): $1.2M · Upper bound (P80): $1.4M · Trigger alert if actual > $1.4M
Decision triggered: "We have a $200K cushion before we hit the budget alarm. Spend the next sprint on growth, not optimisation."
The mistake: "Top 10 most expensive services" — the same boring list every month.
The fix: "Top 10 services that moved this week."
EC2 is always #1. That's not actionable. But SageMaker_endpoint_v3 jumping from #47 to #6 this week is a story worth investigating.
Sort by absolute change (% or $) — not by absolute size.
Decision triggered: "Why did SageMaker triple this week? Open ticket."
The mistake: "Recommendation: Stop unused EBS volumes (potential savings: $4,200/month)" — followed by zero way to act.
The fix: A button right next to the recommendation: [ Apply now ].
The dashboard becomes a control surface, not a viewer. Yes, it's harder to build. Yes, it requires permissions. Yes, it's the difference between a dashboard people use and one they open once a quarter.
Even if "Apply now" just opens a Jira ticket pre-filled with the fix steps — that's still 10x better than "go figure it out yourself".
The mistake: A breakdown by AWS service.
The fix: A breakdown by owner / team / cost centre.
You can't ask "AWS Lambda" to optimise itself. You can ask the Payments team to optimise their Lambda spend. Your dashboard's job is to make ownership unmissable.
If you don't have tagging discipline yet — that is the first FinOps project, full stop.
The mistake: "We detected an anomaly on April 18."
The fix: "We detected the anomaly 2 hours after it started, on April 18."
The KPI of a FinOps team isn't "how much did we save". It's "how fast did we catch the leak". Measure mean-time-to-detection. Plot it as a trend. Your goal is to drive it from days → hours → minutes.
CARTIE AI's internal MTTD is 23 minutes. Most teams' is 8.5 days.
The mistake: A separate ESG dashboard nobody opens.
The fix: kgCO₂e and dollars on the same tile, same chart.
Why? Because the conversation has shifted. CFOs and Sustainability leads now share a budget. Showing them the same data in one place collapses two reviews into one.
April spend: $1.18M · 312.4 tCO₂e (-8% MoM)
Decision triggered: Joint sustainability + finance review actually happens.
While I'm here, let me name the things I see all the time that don't work:
We built the /dashboard on these 8 patterns. Anomaly density, unit economics with cost-per-customer, forecast with bands, top movers, one-click fixes, owner-per-line-item, MTTD trend, and carbon side-by-side.
Not because it looked good in the demo — because every one of those patterns came out of a real conversation where a FinOps lead said "I wish my dashboard did that".
If your FinOps dashboard is failing the "so what?" test, book a 15-minute walkthrough — we'll show you ours and tear yours apart (kindly).
The 8 patterns:
Build any 3 of these and you'll out-pace 90% of FinOps teams running default dashboards.
THE FINOPS BRIEF
Built for finance & engineering teams who are tired of paying for cloud they don't use. No fluff. Just what works.
Unsubscribe anytime. We never sell your data.
ABOUT THE AUTHOR
Founder, CARTIEAI · Building in public
I'm building CARTIE AI to fix the cloud-cost problem I saw drain millions at companies I worked for — where engineering and finance kept talking past each other. If you liked this post, here's where I share unfiltered notes on building this in public: