Ishwar Singh · Analytical Work

Projects &
Case Studies

Structured analysis, financial modelling, and data-driven dashboards — the kind of thinking that sits behind every operational decision.

01 Tier-2 E-Commerce Expansion 02 Academic Intelligence Dashboard
01
Case Study · Financial Modelling

Tier-2 E-Commerce Expansion
— Opportunity vs. Risk

An Indian e-commerce platform plans aggressive expansion into Indore, Lucknow, and Jaipur. Evaluated financial viability across three scenarios — modelling revenue potential, COD fraud exposure, delivery costs, and CAC payback. Designed six operational levers and a phased rollout roadmap.

₹72B
Base Annual Revenue
6M
Monthly Orders
70%
COD Share
5%
Fraud Rate
₹41.67
CAC / Order
80.3%
Contribution Margin
Input Assumptions
ParameterTier-1Tier-2 BaseTier-2 BestTier-2 Worst
Avg Order Value₹1,200₹1,000₹1,200₹800
Monthly Orders10M6M8M4M
COD Share40%70%50%90%
COD Fraud Rate2%5%3%8%
Delivery Cost / Order₹80₹120₹100₹150
CAC₹250₹250₹150₹350
Customer Retention—6 months6 months6 months
Scenario P&L Model
Worst Case
Not justified alone
Monthly Revenue₹3,200M
Annual Revenue₹38,400M
Fraud Loss / mo₹288M
Delivery Cost / mo₹600M
CAC / Order₹58.33
Contribution / Order₹343
Margin42.9%
Best Case
Strong — scale fast
Monthly Revenue₹9,600M
Annual Revenue₹115,200M
Fraud Loss / mo₹144M
Delivery Cost / mo₹800M
CAC / Order₹25.00
Contribution / Order₹908
Margin90.8%
Verdict — Is Year 1 Expansion Financially Justified?

Yes, under Base and Best scenarios. Tier-2 delivers 80.3% contribution margin despite structurally higher delivery cost (₹120 vs ₹80 in Tier-1) and a 2.5× COD fraud rate. The Worst Case remains contribution-positive at 42.9% — meaning even in a downside scenario the unit economics hold. The critical risk is high COD share (90%) compressing margin through fraud. The mitigation priority must be fraud reduction before scale, not after.

Operational Levers — COD & Delivery
01

ML-Based Fraud Scoring

Real-time risk model using pincode fraud history, device fingerprinting, address quality signals, and new vs. repeat customer behaviour. High-risk COD orders flagged for manual review or prepaid nudge.

Target: Fraud rate ≤3% within 90 days of pilot
02

Selective COD Limits

Cap COD at ₹500 for first-time customers in high-fraud pincodes. Unlock progressively based on order completion history. Pair with ₹30–50 prepaid cashback to shift payment mix.

Target: COD share 70% → 55% in 6 months
03

Address Verification at Checkout

OTP-verified address for new COD customers. Google Maps + India Post API to validate deliverability. Reject unserviceable pincodes at order placement, not post-dispatch.

Target: RTO (Return to Origin) rate ≤8%
04

Hub-and-Spoke Micro-Fulfilment

Dark store hubs in high-density Tier-2 zones — Vijay Nagar in Indore, Hazratganj in Lucknow. Last-mile from hub reduces delivery cost 20–30% vs. long-haul origin dispatch.

Target: Delivery cost ≤₹95/order in 12 months
05

Shared Logistics Partnerships

Volume-based agreements with regional 3PLs — Shadowfax, Dunzo, Blowhorn. Bundle delivery with D2C brands entering Tier-2 simultaneously to split fixed infrastructure costs.

Target: 15% reduction in variable delivery cost
06

Repeat Customer Credit Policy

Trust-score system — customers with 5+ successful deliveries and zero fraud get higher COD limits and faster dispatch SLA. Creates behavioural incentive for order completion.

Target: Repeat customer fraud rate ≤1.5%
Phased Rollout Roadmap
Phase 01
Month 1–3 · Pilot

Fraud Foundation

  • Deploy fraud scoring in 2 pilot cities
  • Activate COD caps for new customers
  • Launch address OTP verification
  • Baseline: fraud rate, RTO, COD mix
Fraud <3.5% · RTO <10%
Phase 02
Month 4–6 · Expand

Cost Optimisation

  • Open first micro-hub in top-volume city
  • Sign 3PL partnerships in all 3 cities
  • Launch prepaid cashback incentive
  • Expand fraud model to all cities
Delivery cost ≤₹100 · COD share <60%
Phase 03
Month 7–12 · Scale

Full Rollout

  • Activate repeat-customer credit policy
  • Add 2 additional Tier-2 cities
  • Automate fraud flagging in real-time
  • Annual P&L review vs. benchmarks
Margin ≥75% · Fraud ≤3% · NPS >40
Financial Modelling Scenario Analysis Unit Economics COD Fraud Risk P&L Design Operational Levers Rollout Roadmap Program Management
02
Dashboard · Excel · Data Analysis

Academic Intelligence Dashboard
— Student & Mentor Performance

A multi-sheet Excel performance system tracking 100 students across 10 mentors, 3 streams, and 5 cities. Weighted scoring, z-score relative grading, at-risk early warning, mentor effectiveness rankings, and admission channel correlation — built to drive academic interventions and mentor training decisions.

100
Students
10
Mentors Ranked
3
Streams
5
Cities
4
Channels
6
Months Data
Scoring Methodology

Weighted Final Score

Final score = 0.8 × Academic Score + 0.2 × (Attendance% × 100). Academic component is a weighted average across 4 test types: Unit Test 1 ×0.25, Unit Test 2 ×0.25, Midterm ×0.50, Final ×0.50. Attendance is calculated as total days present ÷ (30 × active months).

Z-Score Relative Grading

Grades assigned by distance from cohort mean (μ) in standard deviations (σ): A = score > μ+σ, B = μ to μ+σ, C = μ−σ to μ, D = μ−2σ to μ−σ, F = below μ−2σ. Normalises performance across streams with different inherent difficulty — prevents grade inflation in easier cohorts.

At-Risk Flag Logic

A student is flagged At Risk if: Cumulative Score < 60 OR Attendance < 10%. Dual-trigger catches both academically struggling students and those disengaging — enabling two distinct interventions: academic support vs. reactivation outreach.

Mentor Effectiveness Formula

Effectiveness = 0.35 × (Avg Score/100) + 0.25 × Avg Attendance + 0.25 × (Satisfaction/100) + 0.15 × (1 − At-Risk%). The final term penalises mentors with high at-risk populations even if raw averages appear acceptable — revealing hidden performance gaps.

Mentor Effectiveness Rankings
Effectiveness Score (0–1) — All 10 Mentors
M03
0.567 ★
M04
0.566
M06
0.516
M10
0.514
M08
0.510
M05
0.508
M02
0.498
M07
0.492⚠ High Risk
M09
0.467
M01
0.442
Mentor Detail — Key Signals
MentorStudentsAvg ScoreAt-Risk %Satisfaction
M03879.612.5%87%
M041077.110.0%95%
M101373.515.4%70%
M061274.225.0%84%
M071072.840.0%82%
M011164.518.2%62%
M07 — Intervention Required. 4 of 10 assigned students are At-Risk — the highest rate across all mentors. Despite a 82% satisfaction score, learning outcomes point to a teaching effectiveness gap requiring immediate structured review.
Student Performance Distribution
6 90+
12 80–89
28 70–79
34 60–69
20 <60
Stream-wise Performance
StreamAvg ScoreAvg AttendanceAt-Risk
Commerce74.278.4%7
Science73.877.1%8
Arts71.573.2%10
At-Risk Students — Sample Flags
IDMentorScoreAttendanceStatus
S065M0469.20%At Risk
S021M0755.360%At Risk
S035M0749.553%At Risk
S001M0678.496.7%Healthy
S052M0291.790%Healthy
Admission Channel Quality Analysis
Avg Weighted Score by Channel
Referral76.4
Direct74.1
Online73.2
Campus71.8
Fee Defaulter Rate by Channel
Referral4.2%
Online7.8%
Direct8.1%
Campus14.3%
Key Insight — Admission Channel

Referral is the highest-quality channel on both dimensions — top weighted scores and the lowest fee-default rate at 4.2%. Campus is the weakest — lowest scores and a 14.3% default rate, more than 3× Referral. This signals a misalignment between campus recruitment messaging and actual programme expectations. Recommendation: increase referral incentives, tighten Campus intake screening, and add a fee-commitment checkpoint at onboarding for Campus admits.

Advanced Excel SUMIFS / COUNTIFS Weighted Scoring Z-Score Grading At-Risk Flagging Cohort Analysis Mentor Ranking Channel Correlation KPI Design Data Storytelling