Cloud Cost Management & FinOps

The accountability gap

Dashboards don't change behavior. ^{Guardrails do.}

Most FinOps programs stall in the same three places. None of them are visibility problems.

The dashboard problem

Reports tell you after the spend.

By the time the chart updates, the money is gone. Last month's overrun isn't actionable — it's a post-mortem.

What Unravel does

Runs cost and performance checks before code ships. Non-compliant deploys don't run.

The expertise problem

Engineers know they should optimize. They don't always know how.

Shift-left has been the answer for a decade. It barely happens — engineers don't have the time, and they don't have the optimization expertise.

What Unravel does

Arvix writes the fix in the PR. Repartition, reorder, resize. Engineers approve the diff — they don't have to invent it.

The accountability problem

You can email a report. You can't make a team own a number.

Without enforced policy, every quarter starts the same: another report, another conversation, another team that's "working on it."

What Unravel does

Cost attributed to job, owner, team. Budgets enforced at workload level. Policy violations don't ship.

——— A new category

Cloud FinOps and _{Data FinOps} aren't the same problem.

Cloud FinOps is a procurement problem — optimize unit costs, allocate spend, negotiate commitments. Data FinOps is an operational problem. The cost is inside the platform, at the query and workload level. You can't negotiate your way out of an inefficient pipeline.

How we're different

You probably already have tools. ^{Here's where they stop.}

Every category solves part of the problem. Unravel is the only one that closes the loop.

——— Works with what you have

Unravel doesn't replace your FinOps stack.
^{It completes it.}

Three questions every FinOps practitioner asks before signing off. Honest answers, not marketing.

Does it account for my commitments and reserved capacity?

Yes. Unravel factors in your existing commitments before surfacing any recommendation. You'll never be told to turn off something you've already paid for.

Does it replace my FinOps tool?

No. It makes your existing tool dramatically more accurate. Unravel pushes workload-level attribution upstream to whatever you're already using for reporting and chargebacks.

How does it prove ROI to the CFO?

Value Realization Tracker shows realized savings vs. target in real time — exportable by BU, team, and pipeline. You walk into the CFO review with a number, not an estimate.

The Data FinOps framework

Prevent. Inform. Optimize. ^Govern.

Four capabilities every mature Data FinOps program requires. What Unravel delivers at each stage.

Catch the cost
before it ships.

Shift-left has been the answer for a decade. It barely happens — engineers don’t have the time, and they don’t have the expertise. Unravel removes both.

CI/CD integration

Cost and performance checks run in the PR, alongside the tests engineers already have. No new tool, no new tab.

Arvix-generated fixes

Inefficient code gets a one-click rewrite — repartition, reorder, resize. Engineers approve the diff, not invent it.

Pre-deploy guardrails

Workloads exceeding budget or violating policy don’t merge. Bad deploys blocked, not reported.

Tag enforcement at source

Untagged jobs flagged before they ship. Chargeback never breaks because attribution is a precondition.

Validated against production

Every Arvix-suggested fix is tested against real workload data — no surprises post-merge.

PR #2847 · feat: revenue-forecast pipeline

Add weekly revenue forecast pipeline

analytics/jobs/revenue_forecast.py · @j.lin

CI · 4 checks

✓ Lintpassed

✓ Unit testspassed

✓ Tag policypassed

△ Unravel · Cost guardrailFAILED

Projected pipeline cost: $87K/mo · +186% over team budget

Full-table scan on transactions — no partition filter. Forecast workload 4× larger than baseline analytics jobs.

Arvix · suggested fix

$87K → $14K/mo

- WHERE date_col >= '2024-01-01' + WHERE date_partition >= '2024-01-01' + AND date_col >= '2024-01-01'

Apply fix · unblock merge

Stop reporting on cost.
Start explaining it.

Most teams can pull a spend chart. Almost none can explain a 40% MoM spike before the CFO asks. Unravel can.

AI Month-over-Month Intelligence

New Entities, Volume Growth, Price Changes, Exited Entities. Every spike explained automatically.

Pipeline-level attribution

Spend by BU, team, and pipeline. Chargeback-ready.

Three-scenario forecasting

Current · Phased · Ideal — through end of FY.

Spend anomaly alerts

Same-hour spike detection across all platforms — before the bill arrives.

Data Temperature

Hot / Warm / Cold storage costs across all platforms, with archival recommendations.

unravel · ai-mom-intelligence · nov

Total spend

$1.24M

+31% vs October

Recoverable

$287K

14 actions queued

New entities

+$94K

3 new pipelines

Volume growth

+$61K

marketing queries

Top cost drivers — explained

analytics_reporting_wh Query regression +$47K

ml_feature_pipeline New entity +$38K

etl_prod_wh Idle hours +$31K

cold_storage · 3 schemas No lifecycle +$22K

From recommendations
to results.

Every other tool tells you what to fix. Unravel fixes it. AutoApply for low-risk actions. Human-in-the-Loop for production code. You set the threshold.

Arvix code rewrites

Query optimizations with side-by-side diffs, validated against real data, one-click apply or rollback.

Warehouse + cluster right-sizing

Sized to actual workload behavior — not guesswork. Across Databricks, Snowflake, and BigQuery.

AutoApply + Human-in-the-Loop

Config changes apply automatically. Code changes need your sign-off. Set per workload.

Storage lifecycle automation

Cold data archived, orphaned tables flagged, time-travel costs managed.

Value Realization Tracker

Realized vs. target savings, MoM ROI — exportable for CFO review.

unravel · optimization-actions

14 actions · $287K/mo recoverable

7 AutoApply ready

⚙

analytics_reporting_wh · Query rewrite

Filter pushdown on date_partition. Full scan → partition scan. Validated vs. 30-day history.

$47.2K/mo

Needs Review

▮

etl_prod_wh · Auto-suspend policy

Suspend at 5 min idle. 14h/day average idle time. No code change.

$31.1K/mo

AutoApply

▣

cold_storage · Archive policy

847 tables, 0 access in 90+ days. Glacier tier. Retention attached.

$34.1K/mo

AutoApply

Stop fixing the same
problem twice.

Optimization without governance means the same issues return next quarter. Governance Decisions prevent waste patterns from recurring — not just flagging them after the fact.

Governance Decisions

Policy rules enforced at the workload level, across all platforms and teams.

Budget controls + anomaly alerts

Per-team thresholds, same-hour spike detection, automated escalation.

Active Governance Agents

Warehouse Sentinel, Unused Cluster Evictor, Stale Table Sweeper — running continuously.

Chargeback-ready reporting

Attribution by BU, team, and pipeline. Realized savings tracked to the dollar.

unravel · governance-decisions · active

12 active policies

All enforcing

◇

No warehouse > XL without approval

Scaling above XL triggers automated review before provisioning.

Active

◷

Auto-suspend dev warehouses at 6PM

47 incidents prevented this month.

Active

▤

Archive tables unaccessed 90+ days

Stale Table Sweeper. Owner notified 14 days before action.

Active

Alert at 80% of monthly team budget

Escalates to VP if no action within 48h.

Active

Trust by design

Every Arvix optimization is validated against real production ^{behavior before it's applied.}

No heuristics. No black boxes. No surprises. If a change would break something, Arvix doesn't apply it. That's why customers let Arvix run on AutoApply for 70%+ of actions.

Zero production incidents from Arvix-applied optimizations · all enterprise customers

The Data FinOps framework

Prevent. Inform. Optimize. ^Govern.

Four capabilities every mature Data FinOps program requires. What Unravel delivers at each stage.

Catch the cost
before it ships.

Shift-left has been the answer for a decade. It barely happens — engineers don’t have the time, and they don’t have the expertise. Unravel removes both.

CI/CD integration

Cost and performance checks run in the PR, alongside the tests engineers already have. No new tool, no new tab.

Arvix-generated fixes

Inefficient code gets a one-click rewrite — repartition, reorder, resize. Engineers approve the diff, not invent it.

Pre-deploy guardrails

Workloads exceeding budget or violating policy don’t merge. Bad deploys blocked, not reported.

Tag enforcement at source

Untagged jobs flagged before they ship. Chargeback never breaks because attribution is a precondition.

Validated against production

Every Arvix-suggested fix is tested against real workload data — no surprises post-merge.

PR #2847 · feat: revenue-forecast pipeline

Add weekly revenue forecast pipeline

analytics/jobs/revenue_forecast.py · @j.lin

CI · 4 checks

✓ Lintpassed

✓ Unit testspassed

✓ Tag policypassed

△ Unravel · Cost guardrailFAILED

Projected pipeline cost: $87K/mo · +186% over team budget

Full-table scan on transactions — no partition filter. Forecast workload 4× larger than baseline analytics jobs.

Arvix · suggested fix

$87K → $14K/mo

- WHERE date_col >= '2024-01-01' + WHERE date_partition >= '2024-01-01' + AND date_col >= '2024-01-01'

Apply fix · unblock merge

Stop reporting on cost.
Start explaining it.

Most teams can pull a spend chart. Almost none can explain a 40% MoM spike before the CFO asks. Unravel can.

AI Month-over-Month Intelligence

New Entities, Volume Growth, Price Changes, Exited Entities. Every spike explained automatically.

Pipeline-level attribution

Spend by BU, team, and pipeline. Chargeback-ready.

Three-scenario forecasting

Current · Phased · Ideal — through end of FY.

Spend anomaly alerts

Same-hour spike detection across all platforms — before the bill arrives.

Data Temperature

Hot / Warm / Cold storage costs across all platforms, with archival recommendations.

unravel · ai-mom-intelligence · nov

Total spend

$1.24M

+31% vs October

Recoverable

$287K

14 actions queued

New entities

+$94K

3 new pipelines

Volume growth

+$61K

marketing queries

Top cost drivers — explained

analytics_reporting_wh Query regression +$47K

ml_feature_pipeline New entity +$38K

etl_prod_wh Idle hours +$31K

cold_storage · 3 schemas No lifecycle +$22K

From recommendations
to results.

Every other tool tells you what to fix. Unravel fixes it. AutoApply for low-risk actions. Human-in-the-Loop for production code. You set the threshold.

Arvix code rewrites

Query optimizations with side-by-side diffs, validated against real data, one-click apply or rollback.

Warehouse + cluster right-sizing

Sized to actual workload behavior — not guesswork. Across Databricks, Snowflake, and BigQuery.

AutoApply + Human-in-the-Loop

Config changes apply automatically. Code changes need your sign-off. Set per workload.

Storage lifecycle automation

Cold data archived, orphaned tables flagged, time-travel costs managed.

Value Realization Tracker

Realized vs. target savings, MoM ROI — exportable for CFO review.

unravel · optimization-actions

14 actions · $287K/mo recoverable

7 AutoApply ready

⚙

analytics_reporting_wh · Query rewrite

Filter pushdown on date_partition. Full scan → partition scan. Validated vs. 30-day history.

$47.2K/mo

Needs Review

▮

etl_prod_wh · Auto-suspend policy

Suspend at 5 min idle. 14h/day average idle time. No code change.

$31.1K/mo

AutoApply

▣

cold_storage · Archive policy

847 tables, 0 access in 90+ days. Glacier tier. Retention attached.

$34.1K/mo

AutoApply

Stop fixing the same
problem twice.

Optimization without governance means the same issues return next quarter. Governance Decisions prevent waste patterns from recurring — not just flagging them after the fact.

Governance Decisions

Policy rules enforced at the workload level, across all platforms and teams.

Budget controls + anomaly alerts

Per-team thresholds, same-hour spike detection, automated escalation.

Active Governance Agents

Warehouse Sentinel, Unused Cluster Evictor, Stale Table Sweeper — running continuously.

Chargeback-ready reporting

Attribution by BU, team, and pipeline. Realized savings tracked to the dollar.

unravel · governance-decisions · active

12 active policies

All enforcing

◇

No warehouse > XL without approval

Scaling above XL triggers automated review before provisioning.

Active

◷

Auto-suspend dev warehouses at 6PM

47 incidents prevented this month.

Active

▤

Archive tables unaccessed 90+ days

Stale Table Sweeper. Owner notified 14 days before action.

Active

Alert at 80% of monthly team budget

Escalates to VP if no action within 48h.

Active

Trust by design

Every Arvix optimization is validated against real production ^{behavior before it's applied.}

No heuristics. No black boxes. No surprises. If a change would break something, Arvix doesn't apply it. That's why customers let Arvix run on AutoApply for 70%+ of actions.

Zero production incidents from Arvix-applied optimizations · all enterprise customers

Not sure where your program stands across these four stages?

The Data FinOps Assessment scores your maturity and shows you exactly where the gaps are.

Take the free assessment

The objection no other FinOps tool answers

Engineering won't fight Unravel.
^{Because Unravel doesn't ask them to learn FinOps.}

The check runs in CI alongside the tests they already have. Arvix writes the fix in the PR — engineers review and approve a diff, the same way they handle any other code change. No new dashboard. No new vocabulary. No new ticket queue.

Engineering owns the cost number because the platform makes ownership the path of least resistance. Not because you mandated it.