Capacity Planning for a Black Friday Traffic Spike

NimbusRetail's checkout service handles roughly 2,000 requests/second on a normal day. Based on last year's Black Friday multiplier and this year's user growth, the infrastructure manager must estimate peak load and required server count — by hand, before touching a calculator app or spreadsheet.

# Aplly.xyz Case Study Submission

## Title
Capacity Planning for a Black Friday Traffic Spike

## Type
Technology/IT

## Difficulty
Intermediate

## Estimated Time
60 minutes

## Overview
NimbusRetail's checkout service handles roughly 2,000 requests/second on a normal day. Based on last year's Black Friday multiplier and this year's user growth, the infrastructure manager must estimate peak load and required server count — by hand, before touching a calculator app or spreadsheet.

## Case Details

Function Focus: Manual estimation, decomposition, back-of-envelope math (no spreadsheet formulas permitted in Phase 1)

Scenario:
NimbusRetail's VP of Engineering needs a server capacity number by tomorrow morning — there's no time to build a full simulation. The infra manager has 11 months of average traffic data, last year's Black Friday peak multiplier, and this year's signup growth rate. They must produce a defensible estimate using structured reasoning alone.

Dataset Structure:
- Month, Average req/sec, Cumulative Signups
- Last year's Black Friday peak req/sec
- This year's YoY signup growth %
- Current per-server capacity (req/sec)
- Current server count

Tasks:
1. Estimate expected peak req/sec for this year's Black Friday using decomposition (base × seasonal multiplier × growth factor)
2. State every assumption explicitly before calculating
3. Compute required server count with a 20% safety margin
4. Identify which single assumption most changes the answer if it's wrong (sensitivity check)
5. Only after submitting your manual answer, re-derive the number using a spreadsheet or AI tool and note any discrepancy

Expected Output:
A one-page estimation memo containing: final server number, full assumptions list, and a sensitivity note flagging the most fragile assumption.

Evaluation Criteria:
Quality of decomposition logic, explicitness of assumptions, reasonableness of final estimate against a reference range, and depth of sensitivity awareness.

## Data Sources

| Month | Avg req/sec | Cumulative Signups |
|---|---|---|
| Jan | 2077 | 40,614 |
| Feb | 1891 | 41,364 |
| Mar | 1836 | 41,968 |
| Apr | 1843 | 43,072 |
| May | 1842 | 43,602 |
| Jun | 1942 | 44,340 |
| Jul | 2171 | 44,867 |
| Aug | 1999 | 46,032 |
| Sep | 2146 | 46,757 |
| Oct | 2261 | 47,541 |
| Nov | 2041 | 48,473 |

Additional Parameters:
- Last year's Black Friday peak: 14,200 req/sec (against a Nov average of ~1,850 req/sec that year)
- This year's YoY signup growth: 35%
- Current capacity per server: 250 req/sec
- Current server count: 10
- Infra budget ceiling: 22 servers max without additional approval

## Solution Frameworks
Fermi estimation, decomposition trees, sensitivity analysis (single-variable "what changes the answer most" check)

## Solver Guidance & Tutorials
Link to: "Estimation Under Uncertainty" tutorial (recommend Aplly author create this as a companion tutorial)

## What You'll Learn
- Structured estimation without relying on tools
- Surfacing and stress-testing hidden assumptions
- Sensitivity analysis for high-stakes, low-time decisions
- Resisting the reflex to reach for AI before reasoning manually

## Tags
capacity planning, estimation, IT operations, Fermi thinking, infrastructure

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