Cloud Cost Runaway Diagnosis

CloudFin Corp's AWS bill jumped sharply in one month. The manager must trace the cause across services and teams using only the raw billing line items — no dashboard, no AI summarizer — before presenting findings to Finance.

# Aplly.xyz Case Study Submission

## Title
Cloud Cost Runaway Diagnosis

## Type
Technology/IT

## Difficulty
Advanced

## Estimated Time
75 minutes

## Overview
CloudFin Corp's AWS bill jumped sharply in one month. The manager must trace the cause across services and teams using only the raw billing line items — no dashboard, no AI summarizer — before presenting findings to Finance.

## Case Details

Function Focus: Manual decomposition of aggregate data, root-percentage calculation, hypothesis testing

Scenario:
Finance wants an explanation by end of day. The bill has 20 line items spanning 4 teams and multiple services. Ranking by percentage change alone is a trap — several small services doubled in percentage terms but barely moved the total dollar figure, while a few large services grew only modestly in percentage but drove most of the actual cost increase.

Dataset Structure:
- Service, Team, Cost This Month ($), Cost Last Month ($), Usage Units This Month, Usage Units Last Month

Tasks:
1. By hand, compute the dollar delta (not percentage delta) per line item and rank contributors — this is the trap most people fall into
2. Identify the top 3 cost drivers by dollar impact and hypothesize a root cause for each
3. For each top driver, distinguish "usage grew proportionally" from "price/config changed disproportionately to usage" — compute cost-per-unit for both months to check
4. Write a one-paragraph executive explanation suitable for a non-technical Finance audience

Expected Output:
Ranked driver list (by $ delta) + root-cause hypothesis for each + one-paragraph executive summary.

Evaluation Criteria:
Correct use of dollar delta over percentage delta for ranking, correct usage-vs-price diagnosis via cost-per-unit calculation, and clarity of the executive summary.

## Data Sources

| Service | Team | Cost Now ($) | Cost Last Mo ($) | Usage Now | Usage Last Mo |
|---|---|---|---|---|---|
| EC2 | Platform | 6,715 | 6,105 | 210,081 | 208,001 |
| S3 | Data | 2,085 | 1,986 | 231,271 | 192,726 |
| RDS | Platform | 2,012 | 1,750 | 20,864 | 20,658 |
| Lambda | Growth | 17,388 | 7,904 | 38,752 | 37,624 |
| CloudFront | Growth | 5,029 | 2,286 | 38,388 | 31,990 |
| ElastiCache | Platform | 3,252 | 2,828 | 397,954 | 331,629 |
| SageMaker | ML | 3,988 | 1,813 | 27,808 | 26,999 |
| EKS | Platform | 1,719 | 1,563 | 168,121 | 152,838 |
| SQS | Growth | 3,321 | 3,163 | 315,339 | 300,323 |
| DynamoDB | Data | 10,976 | 9,979 | 66,036 | 55,030 |
| Redshift | Data | 11,681 | 10,158 | 198,205 | 196,243 |
| API Gateway | Growth | 10,262 | 9,774 | 299,859 | 296,891 |
| VPC NAT Gateway | Platform | 12,582 | 10,941 | 313,516 | 261,264 |
| Glue | Data | 11,707 | 7,805 | 294,130 | 245,109 |
| Kinesis | Growth | 12,336 | 8,224 | 162,908 | 158,164 |
| Athena | Data | 4,306 | 3,745 | 52,698 | 43,915 |
| Step Functions | ML | 17,728 | 5,719 | 482,815 | 459,824 |
| Elastic Beanstalk | Platform | 10,598 | 8,153 | 323,471 | 320,269 |
| Route53 | Platform | 6,014 | 2,734 | 91,861 | 87,487 |
| Secrets Manager | Platform | 10,199 | 3,290 | 224,311 | 222,091 |

(Note: Step Functions, Lambda, Secrets Manager, and Route53 show large $ jumps with near-flat usage — the key diagnostic signal in this case.)

## Solution Frameworks
Manual variance decomposition, cost-per-unit normalization, dollar-impact ranking

## Solver Guidance & Tutorials
Link to: "FinOps Fundamentals: Diagnosing Cost Spikes" tutorial

## What You'll Learn
- Root-cause reasoning on financial/operational data
- Avoiding the percentage-vs-dollar ranking trap
- Communicating technical findings to a non-technical stakeholder

## Tags
FinOps, cloud cost, root cause analysis, data analytics

## Registration Links
- Register as Solver
- Register as Evaluator