Marketplace Supply-Demand Spiral - Pricing Algorithm Backfire

Now I’ll execute the skill with your input: “Generate a hard marketplace RCA with Uber-style feedback focusing on two-sided dynamics”

---

Here’s a complete example investigation demonstrating advanced marketplace RCA skills with Uber-style evaluation:

═══════════════════════════════════════════════════════════════

SCENARIO PRESENTATION

Difficulty: Hard (Marketplace complexity with two-sided dynamics)

You’re the PM for RideShare, a ride-hailing marketplace. Over the past 3 weeks, completed trips have dropped 22% from 2.8M to 2.2M daily trips. The executive team is alarmed about revenue impact and market share loss.

What’s your first question?

═══════════════════════════════════════════════════════════════

INVESTIGATION (Full Conversation)

═══════════════════════════════════════════════════════════════

CANDIDATE: “I need to understand both sides of the marketplace dynamics. Can you show me the trip drop breakdown by rider segments AND driver supply metrics including active drivers, online hours, and earnings?”

INTERVIEWER: “Excellent opening - immediately thinking about both sides! This is exactly what we look for at Uber.

Rider-side breakdown (completed trips):

• Daily commuters (10+ trips/month): 1.4M → 1.3M (-7%)
• Regular users (3-9 trips/month): 900K → 750K (-17%)
• Occasional users (1-2 trips/month): 500K → 150K (-70%)

Driver-side metrics:

• Active drivers (drove at least 1 trip): 85K → 73K (-14%)
• Average online hours per driver: 6.2 hrs/day → 5.1 hrs/day (-18%)
• Driver hourly earnings: $28/hr → $22/hr (-21%)
• Driver utilization (time with passenger): 58% → 47% (-19%)

What patterns are you seeing here?”

═══════════════════════════════════════════════════════════════

CANDIDATE: “The drop is heavily concentrated on occasional users (-70%), and driver supply is collapsing across multiple dimensions. Lower driver earnings and utilization suggest this started on the supply side. When drivers earn less and spend more time idle, they leave or reduce hours, which impacts service quality. Can you show me wait times, surge pricing frequency, and cancellation rates?”

INTERVIEWER: “Strong hypothesis about supply-side trigger causing demand impact! Here’s the data:

Service quality metrics:

• Average wait time: 4.2 min → 7.8 min (+86%)
• Trips with >10 min wait: 12% → 38%
• Surge pricing frequency: 18% of trips → 32% of trips
• Driver cancellation rate: 4% → 11%
• Rider cancellation rate: 6% → 15%

Geographic breakdown (trip volume change):

• Dense urban (Manhattan, SF Downtown): -12%
• Suburban: -28%
• Airport: -18%

What’s the feedback loop you’re identifying?”

═══════════════════════════════════════════════════════════════

CANDIDATE: “This is a vicious cycle. Fewer drivers → longer wait times and more surge pricing → occasional users drop off first because they’re price-sensitive and won’t wait → lower trip volume → drivers earn less per hour because they’re idle more → more drivers leave or reduce hours → even worse service. The suburban drop is worse because driver density is already lower there, so any supply reduction has outsized impact. Can you show me what changed 3 weeks ago that could have triggered the initial driver supply reduction?”

INTERVIEWER: “Exactly right on the reinforcing feedback loop! Here’s what changed:

Policy change 3 weeks ago:
We implemented a new dynamic pricing algorithm to reduce rider wait times during non-peak hours. The algorithm lowered base fares by 15% during off-peak periods (10am-4pm, 8pm-12am) and increased surge multipliers during peak.

Impact by time period:

• Peak hours (7-9am, 5-7pm):

  • Driver earnings: $32/hr → $33/hr (+3%)
  • Trips: 45% of daily volume (stable)

• Off-peak hours (rest of day):

  • Base fare: $8.50 → $7.20 (-15%)
  • Driver earnings: $24/hr → $18/hr (-25%)
  • Trips: 55% of daily volume → 48% of daily volume (-13%)

Driver behavioral changes:

• Drivers who primarily work off-peak: 42K → 29K (-31%)
• Drivers switching to competitors: ~8K (11% of driver base)

What’s the complete root cause and causal mechanism?”

═══════════════════════════════════════════════════════════════

CANDIDATE: “The root cause is the dynamic pricing algorithm optimization that backfired by destroying driver economics during off-peak hours, which represents 55% of daily volume.

Complete causal mechanism:

1. Algorithm change reduced off-peak base fares by 15% to stimulate demand
2. Driver economics collapse: Off-peak earnings dropped from $24/hr to $18/hr (-25%), making it uneconomical compared to minimum wage and alternative gig platforms
3. Supply exodus: 31% of off-peak drivers left (13K drivers), and 8K drivers switched to competitors where earnings remained stable
4. Service degradation: Remaining drivers can’t cover demand, causing wait times to jump from 4.2 to 7.8 minutes (+86%) and surge frequency to double to 32%
5. Demand destruction: Occasional users (-70%) and regular users (-17%) churn because of poor experience (long waits, high surge prices), while daily commuters are more captive
6. Feedback loop intensifies: Lower trip volume reduces driver utilization from 58% to 47%, cutting hourly earnings further even during hours that weren’t directly affected, causing more drivers to reduce hours or leave
7. Geographic amplification: Suburban areas suffer more (-28%) because driver density is lower, so supply reduction creates service deserts

Quantified impact:

• 12K drivers lost × 6 hrs/day × 2.5 trips/hr = 180K trips/day lost from direct supply reduction
• Additional 420K trips lost from induced demand destruction via poor service quality
• Total: 600K daily trips lost (matches the 2.8M → 2.2M drop)

The algorithm optimized for a local maximum (rider wait times) but created a global minimum by breaking the marketplace equilibrium.”

INTERVIEWER: “Outstanding! That’s a complete marketplace analysis with quantified feedback loops and multi-order effects. You demonstrated senior PM-level thinking. Let me provide your evaluation.

---

PERFORMANCE EVALUATION

Overall Score: 4.9/5 - Outstanding (Staff/Principal PM level)

What You Did Well:

✅ Two-sided thinking from first question - Immediately requested both rider segments AND driver metrics (supply, earnings, utilization) in a single bundled question
✅ Identified feedback loop hypothesis early - Recognized supply-side trigger before seeing service quality data
✅ Connected geographic patterns to driver density - Showed deep marketplace intuition that sparse markets amplify supply shocks
✅ Traced complete causal chain - Algorithm → driver economics → supply exodus → service degradation → demand destruction → reinforcing loop with 7 explicit steps
✅ Quantified impact at each stage - Specified $6/hr earnings drop, 31% off-peak driver loss, 86% wait time increase, 70% occasional user churn
✅ Calculated trip math - Worked backwards from driver loss to estimate trip impact (180K direct + 420K induced = 600K total)
✅ Strategic insight - Identified local vs global optimization tradeoff and marketplace equilibrium concept

Areas for Improvement:

⚠️ Recovery estimation - Could have asked “If we revert the pricing change, how quickly can we recover the 12K lost drivers and 600K trips?”
⚠️ Competitive dynamics - Could have asked earlier about competitor actions: “Did Lyft or other platforms change pricing 3 weeks ago?”

Uber-Style Calibration:

Two-sided marketplace dynamics: ✓✓✓ Exceptional

• Looked at both sides in opening question
• Understood supply-demand interaction at each step
• Recognized that both sides reinforce each other in a downward spiral

Feedback loops and cascading effects: ✓✓✓ Exceptional

• Identified vicious cycle before seeing full data
• Traced 7-step causal chain from algorithm to trip loss
• Understood second-order effects (lower volume → worse utilization → further earnings drop)

Impact quantification: ✓✓ Strong

• Quantified each step with specific percentages
• Calculated trip loss from driver exodus
• Could strengthen by estimating revenue impact ($X million/week)

External competitive factors: ✓ Good

• Identified 8K drivers switching to competitors
• Could have probed competitor pricing changes earlier

Geographic and segment insight: ✓✓ Excellent

• Connected suburban amplification to driver density
• Understood occasional users churn first due to price sensitivity

Uber readiness: 98% - You’re ready for Staff+ marketplace roles at Uber, Lyft, DoorDash, Airbnb

Key Learnings:

Marketplace RCA Pattern You Demonstrated:

1. Open with both sides: Always request supply AND demand metrics in first question
2. Look for economics: Check earnings, utilization, and incentives for each participant
3. Identify trigger side: Determine if supply caused demand drop or vice versa
4. Trace feedback loops: Map how one side’s behavior impacts the other, which reinforces the first side’s behavior
5. Quantify cascade: Estimate impact at each step (direct effect → induced effect → reinforcing effect)
6. Consider geography: Dense vs sparse markets react differently to supply/demand shocks
7. Segment by sensitivity: Occasional users churn first, power users are more captive

Why This Matters:
Marketplaces fail when equilibrium breaks. Unlike single-sided products where you optimize for one user, marketplaces require maintaining balance. Over-optimizing for one side (riders) without considering the other side (drivers) can destroy the entire system through feedback loops.

Next Steps:

You’re interview-ready, but to reach 100%:

• Practice recovery estimation: “What % of lost supply/demand can we recover in 1 week, 1 month, 1 quarter?”
• Add revenue impact: Convert trip loss to revenue impact including surge pricing effects
• Practice competitor analysis: Always ask about competitive moves as potential external triggers
• Try three-sided marketplace (e.g., Uber Eats with restaurants, drivers, eaters)

Recommended practice:

• Real interviews: You’re ready
• Advanced practice: Three-sided marketplace RCA (UberEats, Airbnb Experiences)
• Speed practice: Aim for this same depth in 12-15 minutes instead of 18-20 minutes

═══════════════════════════════════════════════════════════════

Study Guide for This Example

Framework Applied:

1. Both sides immediately (rider segments + driver metrics)
2. Identify worse-affected side (drivers: -14% active, -21% earnings)
3. Check service quality (wait time +86%, cancellations up)
4. Connect timing to changes (pricing algorithm 3 weeks ago)
5. Trace mechanism: Change → Economics → Behavior → Service → Demand → Reinforcement
6. Quantify each step
7. Explain why certain segments affected more (occasional users, suburban)

Key Data Requests That Unlocked the Case:

1. Opening: Both rider segments AND driver supply/earnings
2. Follow-up: Wait times, surge frequency, cancellations
3. Timing: What changed 3 weeks ago
4. Deep dive: Off-peak vs peak impact breakdown

What Made This a 4.9/5:

• Bundled questions efficiently (both sides in opening)
• Stated hypotheses explicitly with reasoning
• Quantified impact at every step
• Identified feedback loops before full data revealed
• Explained WHY geography matters (driver density)
• Connected algorithm change to participant behavior to market failure

To Reach 5.0/5:
Ask about recovery timeline and estimate revenue impact in dollars, not just trips.