Interview Vault

Step 1: Clarify the problem
- Which airline segment? (Commercial passenger, cargo, loyalty?)
- Geographic scope? (Domestic, international?)
- Timeframe for turnaround — 18 months is aggressive

Step 2: Structure (Profitability Framework)
Profit = Revenue - Costs
Revenue = Passengers × Price per ticket + Ancillary (bags, upgrades)
Costs = Fuel + Labor + Aircraft + Overhead

Step 3: Identify key drivers to investigate
Revenue side:
- Load factor (empty seats = lost revenue)
- Price per seat (yield management)
- Ancillary revenue (bags, meals, premium cabin upgrades)

Cost side:
- Fuel (largest variable cost, hedging strategy)
- Labor (unionized, pensions)
- Aircraft financing (owned vs leased)

Step 4: Hypothesis-driven analysis
Likely culprits:
- Fuel prices spiked 40% in last 2 years
- Discount competitors (Spirit, Frontier) stealing market share
- Legacy pension obligations

Step 5: Recommendation
- Short-term: Fuel hedging + capacity reduction on unprofitable routes
- Medium-term: Dynamic pricing + loyalty program upsell
- Long-term: Fleet modernization + pension restructuring

Step 1: Clarify
- Which luxury brand? (BMW, Mercedes, Porsche?)
- Price point for EV? (Premium $80K+ or mid-tier $50K?)
- Timeframe to launch?

Step 2: Market Attractiveness
- Market size: US EV market ~1.4M units (2023), growing 40% YoY
- Luxury EV segment: ~250K units, ~$25B
- Tesla dominates (60% share), followed by Rivian, Lucid

Step 3: Competitive Position
- Strengths: Brand, dealer network, engineering expertise
- Weaknesses: Behind on EV battery tech, no charging network
- Threats: Chinese EVs (BYD, NIO) expanding globally

Step 4: Entry Strategy Options
- Option A: Build from scratch (control, slow, high capex)
- Option B: Partner with existing OEM (speed, less control)
- Option C: Acquire startup (Lucid, Fisker — fast but risky)

Step 5: Recommendation
- Enter with premium SUV ($80K+)
- Partner with LG Chem for batteries, Electrify America for charging
- Launch in CA, TX, FL (top EV adoption states)
- Target 50K units/year by 2027

Step 1: Clarify
- What industry vertical? (Vertical SaaS has higher multiples)
- Current revenue multiple? (5x = $500M / $100M ARR)
- Growth rate: 30% YoY is good but not elite (elite = 50%+)
- Rule of 40? Need to know EBITDA margin

Step 2: Market Analysis
- TAM for this SaaS vertical?
- Market growth rate (should exceed 15%)
- Competitive landscape (fragmented or dominated by incumbent?)

Step 3: Company Analysis
- Net Revenue Retention (NRR)? Target >110%
- Customer concentration (top 10 customers % of revenue)
- Logo churn vs revenue churn
- Sales efficiency (CAC payback period)

Step 4: Financials
- EBITDA margin at exit (target 25-35%)
- Rule of 40: Growth + EBITDA margin
- Exit multiple assumption (6-8x revenue typical)

Step 5: Deal Structure
- Entry multiple (5x) vs exit multiple (6-8x)
- Expected IRR if exit in 5 years at $800M-$1.2B
- Leverage ratio and interest coverage

Step 6: Recommendation
PROCEED IF:
- NRR >110%, logo churn <5%
- Rule of 40 score >40
- TAM growing 15%+

WALK IF:
- Customer concentration >30% in top 10
- Rule of 40 <30
- Platform cannot scale without major re-platforming

Step 1: Clarify
- Residential vs commercial chargers?
- Including Tesla Superchargers or only public?
- Define charging station (level 2 vs DC fast charging?)

Step 2: Top-down approach
- 2030 US EV fleet projection: 50M vehicles (based on 20% new sales EV + replacement)
- Charging ratio: 1 public charger per 10 EVs (industry benchmark)
- US needs: 50M / 10 = 5M public chargers

Step 3: Bottom-up validation
- Daily EV charging demand per vehicle: 30 kWh
- Fast charger output: 150 kW, serves 5 cars/day
- Chargers needed: 50M × 0.2 (public charging share) / 5 = 2M fast chargers
- Plus 3M Level 2 chargers in commercial/workplace locations

Step 4: Reconcile and sanity check
- Total: ~5M chargers across levels
- Current: ~150K public chargers (2024)
- Growth needed: 33x in 6 years = ~70% CAGR

Step 5: Answer
US will need ~5M EV charging stations by 2030, requiring $100B+ in infrastructure investment.

Multiple valid answers (interviewer wants structured thinking):

1. Safety (most common answer):
   - Round covers can't fall through the hole regardless of orientation
   - Square/rectangular covers can fall diagonally through the opening

2. Manufacturing:
   - Easier to cast and machine round shapes
   - Standardized sizes work across all cities

3. Transport:
   - Can be rolled when heavy (300+ lbs)
   - Easier to position and align

4. Pressure distribution:
   - Round shape distributes pressure evenly
   - Resists cracking under vehicle weight

STRONG STAR STRUCTURE:

Situation (10-15 sec):
"During my junior year, I was elected president of our 40-person CS club. In October, our main hackathon event — scheduled for 500+ attendees — was canceled 2 weeks before due to venue issues."

Task (5-10 sec):
"As president, I needed to either cancel the hackathon (letting down 500 students, losing $15K in sponsorships) or find a new venue in 14 days."

Action (30-40 sec — most important):
"I immediately split my 10-person exec team into 3 workstreams:
1. I personally cold-emailed 25 campus departments and found an available auditorium.
2. I negotiated with sponsors (Google, Microsoft) to maintain funding despite the chaos.
3. I rebuilt the event logistics in 48 hours — new signage, catering, judges.
4. I over-communicated with attendees via 3 email updates to prevent dropouts."

Result (10-15 sec):
"We hosted the hackathon on schedule with 95% of registered attendees showing up. We secured $5K in additional sponsorship from Google due to how we handled the crisis. The club was invited to speak at a leadership summit about crisis management."

STRONG EXAMPLE (Pick a real, instructive failure):

Situation: "In sophomore year, I led a startup project to build a dorm food delivery app. We had 500 users at peak."

Task: "I was CEO, responsible for strategy, hiring 3 co-founders, and investor pitches."

Action (what went wrong — be HONEST):
"I made 3 critical mistakes:
1. I ignored user feedback asking for refunds when restaurants delivered late. I prioritized adding features over fixing the core experience.
2. I under-paid my delivery drivers, leading to high turnover and worse service.
3. I spent $8K on marketing before product-market fit, burning capital quickly."

Result:
"Within 6 months, user churn hit 70% and we shut down the project. I personally lost $5K of my savings. However, I learned 3 lessons that fundamentally changed how I build products:
1. Fix bugs > add features (my current project has 95% retention vs 30%)
2. Pay fair wages to attract/retain quality team members
3. Validate product-market fit BEFORE scaling spend"

STRONG STAR:

Situation: "During my software engineering internship at [Company], I was pair-programming with a senior engineer who insisted we use a complex design pattern (Factory Pattern) for a simple feature."

Task: "I believed we should use a straightforward approach to save development time, but the senior engineer disagreed strongly."

Action:
"I didn't just push my opinion. Instead:
1. I prototyped BOTH approaches in a side branch (2 hours)
2. I invited the senior engineer to code review each
3. I presented data: simple approach = 150 lines, Factory = 400 lines
4. I asked thoughtful questions: 'What future flexibility does the Factory give us that we'll actually need?'
5. We discussed openly, and he realized the complexity wasn't justified"

Result:
"We shipped the simple version. 6 months later, no additional complexity was needed. The senior engineer thanked me for pushing back thoughtfully and invited me to help design the next feature architecture."

STRONG STAR:

Situation: "During my PM internship at [Startup], my manager asked me to 'improve user engagement' with no specific metric or timeline."

Task: "I had 2 weeks to deliver a plan and start execution — but 'engagement' could mean DAU, session length, feature usage, retention, etc."

Action:
"I didn't just pick a metric arbitrarily. I:
1. Scheduled 1:1s with 5 stakeholders (CEO, data, design, engineering, support) to understand what each cared about
2. Analyzed historical data: our DAU/MAU had dropped from 45% to 35% over 6 months
3. Hypothesized the root cause: new user onboarding was too complex (62% drop-off at step 3)
4. Proposed a specific metric to improve: increase 7-day retention from 25% to 35%
5. Shipped 3 experiments: onboarding simplification, email nudges, personalized recommendations"

Result:
"Within 6 weeks, 7-day retention improved from 25% to 38%. DAU/MAU recovered to 42%. My manager said my ability to turn ambiguous asks into specific plans was the single most valuable skill I brought."

STRUCTURE: 3 specific reasons tied to your goals.

STRONG EXAMPLE (Google):
"Three reasons:

1. SCALE: I want to build products that impact billions of users. When I worked on my dorm food delivery app with 500 users, I realized the complexity of scaling. Google Maps serves 2B+ monthly users — the engineering challenges at that scale fascinate me.

2. CULTURE OF ENGINEERING EXCELLENCE: I've read engineering blog posts by Jeff Dean and Sanjay Ghemawat on MapReduce and Bigtable. Google's engineering culture — where you can propose a design doc that reshapes an entire system — is where I want to grow.

3. SPECIFIC TEAM INTEREST: I'm particularly interested in the Google Pay India team. Having grown up in India, I've seen how UPI transformed everyday life. The chance to help millions of merchants accept payments and grow their businesses is deeply meaningful to me.

Why NOT just FAANG as a brand? Because I've researched specific teams, people, and problems I want to solve at Google specifically."

STRUCTURE (90 seconds):

Past (20 sec):
"I grew up in [City] and discovered my passion for software engineering in high school when I built a tutoring app that helped 200 students at my school."

Education + Why CS (20 sec):
"I'm a senior at [University] studying Computer Science with a focus on ML. I chose this path because I wanted to build systems that solve real problems — not just write code."

Key Experience 1 (20 sec):
"My most formative experience was my internship at Stripe last summer. I owned a project to reduce payment processing errors by 15% through improved retry logic. This taught me how backend systems scale."

Key Experience 2 (20 sec):
"Before that, I co-founded a dorm food delivery startup. We hit 500 users in 6 months before I decided to focus on honing deeper technical skills. It taught me product-market fit matters more than features."

Future / Why Now (10 sec):
"I'm now looking for a new grad SWE role at [Company] where I can combine my technical growth with product thinking. I'm particularly excited about the [Team] because of [specific reason]."

APPROACH 1: Brute Force O(n²)
for i = 0 to n:
  for j = i+1 to n:
    if nums[i] + nums[j] == target: return [i, j]

Time: O(n²), Space: O(1)

APPROACH 2: Hash Map O(n) [OPTIMAL]
seen = {}
for i = 0 to n:
  complement = target - nums[i]
  if complement in seen: return [seen[complement], i]
  seen[nums[i]] = i

Time: O(n), Space: O(n)

COMMUNICATION TIPS:
- Clarify: Are there duplicates? Can we assume exactly one solution?
- Mention brute force first, then optimize
- Walk through example: nums=[2,7,11,15], target=9
  - i=0: seen={}, complement=7, add 2→0. seen={2:0}
  - i=1: complement=2, found! return [0, 1]
- Discuss edge cases: empty array, no solution exists

APPROACH: Stack-based

Algorithm:
1. Create a stack and a mapping: {')': '(', ']': '[', '}': '{'}
2. Iterate through string:
   - If character is opening bracket, push to stack
   - If closing bracket, check if stack top matches
     - If matches: pop
     - If not: return false
3. Return stack.isEmpty()

Code (Python):
def is_valid(s):
    stack = []
    pairs = {')': '(', ']': '[', '}': '{'}
    for char in s:
        if char in '([{':
            stack.append(char)
        else:
            if not stack or stack.pop() != pairs[char]:
                return False
    return len(stack) == 0

Time: O(n), Space: O(n)

EDGE CASES:
- Empty string → true
- Only opening brackets '(((' → false
- Only closing brackets ')))' → false
- Mismatched '([)]' → false

ITERATIVE APPROACH:

def reverse_list(head):
    prev = None
    curr = head
    while curr:
        next_node = curr.next
        curr.next = prev
        prev = curr
        curr = next_node
    return prev

Time: O(n), Space: O(1)

RECURSIVE APPROACH:

def reverse_list(head):
    if not head or not head.next:
        return head
    new_head = reverse_list(head.next)
    head.next.next = head
    head.next = None
    return new_head

Time: O(n), Space: O(n) — call stack

VISUAL (Iterative):
Original: 1 → 2 → 3 → 4 → None
Step 1: prev=None, curr=1, next=2
Step 2: curr.next = prev → 1 → None, prev=1, curr=2
Step 3: 2 → 1 → None, prev=2, curr=3
Step 4: 3 → 2 → 1 → None, prev=3, curr=4
Step 5: 4 → 3 → 2 → 1 → None
Return prev (new head) = 4

WHAT IS OVERFITTING:
A model performs well on training data but poorly on unseen (test) data. The model has memorized noise/patterns specific to training set rather than learning generalizable patterns.

HOW TO DETECT:
1. Train/validation loss gap: Training loss keeps decreasing while validation loss increases
2. High training accuracy, low test accuracy (e.g., 99% train, 70% test)
3. Learning curves show large gap between train/val performance

PREVENTION TECHNIQUES:

1. More Data:
   - Collect more training examples
   - Data augmentation (image rotations, text paraphrasing)

2. Regularization:
   - L1 (Lasso): adds |w| to loss, drives weights to zero (feature selection)
   - L2 (Ridge): adds w² to loss, keeps weights small
   - Dropout (neural nets): randomly zero out neurons during training

3. Model Complexity:
   - Reduce model size (fewer parameters)
   - Early stopping: halt training when val loss stops improving

4. Cross-Validation:
   - K-fold cross-validation to get robust performance estimates

5. Ensemble Methods:
   - Random Forests, Gradient Boosting (combine many weak learners)

EXAMPLE SCENARIO:
"If I trained a decision tree to depth 50 on 1000 samples, it would likely overfit. I'd:
- Prune the tree (max_depth=10)
- Apply ensemble (Random Forest)
- Use cross-validation to tune hyperparameters"

APPROACH: Normal Approximation to Binomial

Setup:
- X = number of heads in 100 flips
- X ~ Binomial(n=100, p=0.5)
- Mean μ = np = 50
- Variance σ² = np(1-p) = 25
- Std Dev σ = 5

For large n (n=100 is large enough), Binomial ≈ Normal(μ, σ²)

Find P(X > 60):
- Standardize: Z = (60 - 50) / 5 = 2
- P(X > 60) = P(Z > 2) ≈ 0.0228 (from standard normal table)

ANSWER: ~2.28% or ~1/44

QUICK MENTAL MATH:
- Z=1 → P(Z>1) ≈ 16%
- Z=2 → P(Z>2) ≈ 2.5%
- Z=3 → P(Z>3) ≈ 0.1%

INTERPRETATION:
Even for a fair coin, getting 60+ heads out of 100 is quite rare (<2.5%). This is why 95% confidence intervals are typically ±1.96 std devs from mean.

CONTINUITY CORRECTION (bonus):
For more accuracy: P(X > 60) = P(X ≥ 60.5) → Z = 2.1, P ≈ 1.79%

APPROACH 1: Geometric Distribution Formula

X ~ Geometric(p=1/6)
E[X] = 1/p = 1/(1/6) = 6

ANSWER: 6 rolls on average.

APPROACH 2: Derivation from First Principles (if interviewer asks)

Let E = expected number of rolls to get a 6.

On each roll:
- With probability 1/6: get a 6 (1 roll total)
- With probability 5/6: don't get a 6 (1 roll + E more rolls)

E = (1/6)(1) + (5/6)(1 + E)
E = 1/6 + 5/6 + 5E/6
E = 1 + 5E/6
E - 5E/6 = 1
E/6 = 1
E = 6

INTUITIVE CHECK:
Over many trials, you get a 6 about 1/6 of the time, so you need about 6 trials to see one.

FOLLOW-UP (variance):
Var(X) = (1-p)/p² = (5/6)/(1/36) = 30
Std Dev ≈ 5.48

STEP 1: COMPREHEND THE PROBLEM

Clarifying questions:
- What's the goal? User engagement, revenue, creator retention?
- Which user segment? Creators or viewers?
- Timeframe? MVP in 3 months or 1-year roadmap?
- Any constraints? (Technical, legal, resource?)

Assume: Goal is to increase creator retention on YouTube Shorts (currently losing creators to TikTok).

STEP 2: IDENTIFY USER PERSONAS

1. Casual Creator (makes <5 videos/month): Needs easy editing tools
2. Professional Creator (full-time, 20+ videos/month): Needs monetization + analytics
3. Aspiring Creator (10-500 followers): Needs discoverability + growth tools

Focus on: Professional Creators (biggest ARR impact)

STEP 3: REPORT USER NEEDS

Pain points for Professional Creators:
- YouTube Shorts monetization is weak vs TikTok Creator Fund
- Analytics are shallow compared to TikTok
- Short-form videos don't convert to long-form watch time
- Tedious cross-posting from TikTok

STEP 4: CUT (PRIORITIZE)

Top 3 problems to solve:
1. [HIGH] Monetization: Pay creators competitive rates
2. [HIGH] Analytics: Deep creator insights (retention curves, demographics)
3. [MEDIUM] Cross-platform publishing: Import TikTok videos easily

STEP 5: LIST SOLUTIONS

For "Monetization":
- Option A: YouTube Shorts Creator Fund (fixed pool, like TikTok)
- Option B: AdSense revenue share on Shorts (similar to long-form videos)
- Option C: Branded content marketplace (connect creators with brands)
- Option D: Super Thanks / Super Chat for Shorts

RECOMMENDED: Option B — AdSense share
- Aligns with YouTube's existing model (creators already understand)
- Rewards view quality, not just viral spikes
- Scales with YouTube's ad business

STEP 6: EVALUATE

Pros of AdSense Share:
- Predictable revenue for creators
- Rewards engagement + ad quality
- Reduces complexity (one monetization model across YouTube)

Cons:
- Lower per-view CPM for Shorts (lower-quality ads)
- Creators used to higher Creator Fund payouts might be unhappy

Mitigation: Launch with 55/45 split (vs 55/45 on long-form), guarantee minimum $0.10/1000 views.

STEP 7: SUMMARIZE

"To compete with TikTok, I'd prioritize creator monetization via AdSense revenue share. This leverages YouTube's existing infrastructure, aligns incentives with ad quality, and gives creators the predictable income they need. I'd launch with a competitive 55/45 split and guaranteed minimum CPM. Success metric: Monthly Active Creators uploading >10 Shorts/month should grow 30% in 6 months."

STEP 1: CLARIFY

Questions:
- Which market? US, India, Latin America?
- What's the primary business goal? Growth, retention, profitability?
- Any specific user segment? Casual eaters, families, business orders?

Assume: US market, goal is to increase repeat ordering (retention).

STEP 2: UNDERSTAND THE 3-SIDED MARKETPLACE

Supply Side: Restaurants + Drivers
- Restaurants: Want more orders, lower commission
- Drivers: Want predictable earnings, fewer wait times

Demand Side: Eaters
- Want: Faster delivery, lower fees, better food quality

Platform: Uber
- Wants: Higher order frequency, better unit economics

STEP 3: IDENTIFY TOP RETENTION ISSUES

Pain points causing lost retention:
1. Delivery takes too long (45+ min for simple orders)
2. Hidden fees (service fee + delivery fee + small order fee)
3. Food arrives cold or damaged
4. Hard to re-order favorite meals quickly

STEP 4: PRIORITIZE

Top 3 by impact + effort:
1. [HIGH IMPACT, MEDIUM EFFORT] "Reorder" shortcut on home screen
2. [HIGH IMPACT, HIGH EFFORT] Guaranteed delivery times (refund if late)
3. [MEDIUM IMPACT, LOW EFFORT] Transparent fee breakdown (reduce surprises)

STEP 5: DEEP DIVE ON "GUARANTEED DELIVERY"

Feature: "Uber Eats Promise — Delivery in 30 min or 30% off"

Design:
- Show estimated delivery time BEFORE ordering (not after)
- If delivery exceeds estimate by 10+ min, auto-refund 30%
- Use ML to predict delivery times accurately (consider restaurant prep, traffic, driver availability)
- Only offer on orders where confidence is >90%

Business Impact:
- Reduces customer complaints by 25% (data: late deliveries = #1 complaint)
- Increases order frequency: eaters who trust delivery times order 2x more
- Costs: Refunds (~2-3% of orders), but offset by retention gains

Metrics to Track:
- Primary: 30-day retention rate (goal: +15%)
- Secondary: Avg orders per month per user (goal: +20%)
- Guardrail: Refund rate (keep <5%)

STEP 6: RECOMMENDATION

"Uber Eats Promise" would be my top recommendation because:
1. Addresses the #1 pain point (late deliveries)
2. Builds trust that drives retention
3. Measurable financial impact (+15% retention = $500M+ ARR increase)
4. Technical feasibility (we have the data to predict accurately)