Whatnot Interview Prep

Commerce Strategy & Operations — Round 2 Live Case

$8B
GMV (2025) [↗]
$11.5B
Valuation [↗]
~$1B
Revenue (est.)
250+
Categories
9
Countries
20M+
New Accounts

Interview Format

Format 30 min total
Structure 5-10 min intro → Live brainstorming case
What they test Problem approach, questions you ask, data you'd look at
Interviewer Ops team member (not recruiter)
Beauty +791% Electronics +444% Jewelry +259% Fashion +223% [↗]

Company Intelligence

299 jobs tracked • 214 active • Data from Trackly LAMP Monitor

Hiring by Function

Engineering
99 (33%)
Other
57 (19%)
People/HR
31 (10%)
Operations
22 (7%)
Support/CS
20 (7%)
Strategy
15 (5%)
Marketing
11 (4%)
Data/Analytics
11 (4%)
Product
8 (3%)
Sales
8 (3%)

Top 7 Operational Priorities

🚛

Logistics & Shipping CRITICAL

3x Sr Manager hiring, building shipping infrastructure at unprecedented scale, multi-carrier management

🌍

International Expansion HIGH

Europe grew 600%, France: fastest-growing European market, hiring in Dublin/London/Berlin/Tokyo/Sydney, localized trust ops

🛡

Trust, Risk & Fraud HIGH

ML fraud detection, policy violation systems, dispute resolution, dedicated Data & Marketplace Integrity team

📊

Data Infrastructure HIGH

$230-290K salary band, Snowflake/Kafka/Flink/dbt stack, retrieval platforms, real-time streaming

💳

Payments & Commerce MEDIUM

Cross-border payments, tax compliance, payout timing, recently filled several key roles

🎧

Customer Experience MEDIUM

20+ CX roles, high-value item specialists, overnight shifts, 24/7 support scaling

📈

Rapid Ops Scaling MEDIUM

"Move remarkably fast with little structure", dashboards, marketing ROI, AI tools fluency required

Geographic Footprint

SF NYC LA Phoenix Seattle London Dublin Berlin Tokyo Sydney Krakow China

Tech Stack

Elixir/Phoenix Go Kafka Redis/Valkey Snowflake dbt Agora GraphQL OpenSearch Terraform

User Personas

Buyer & seller archetypes driving Whatnot's marketplace dynamics

Buyer Personas

The Collector
Core Power User
Age: 25-40 • 70% male • $50K-$90K income
Engagement: ~95 min/day [↗], 16 orders/year (Gen Z), 80%+ retention
Categories
Sports cards, Pokemon TCG, Funko, coins
Motivation
Thrill of the hunt, nostalgia + investment, community identity
Pain Points
Bank flags multiple charges, grading inconsistency, mystery box gambling feel
The Deal Hunter
Growth Segment
Age: 22-35 • 55% female • $30K-$60K income
Engagement: 1-2x/week purchases, $15-$50 AOV
Categories
Fashion, beauty, sneakers, jewelry, electronics
Motivation
Getting deals on brands, entertainment + shopping combined
Pain Points
Quality inconsistency, hard returns, overwhelmed by choices
The Entertainment Watcher
Emerging Segment
Age: 18-28 • 50/50 gender • $20K-$45K income
Engagement: Watches 10 streams per 1 purchase, huge chat engagement
Categories
Whatever's trending
Motivation
Entertainment, social interaction, FOMO
Why They Matter
Creates "audience liquidity" — viewers drive bids higher
The Reseller/Flipper
High-Value Niche
Age: 22-45 • 75% male • Variable income
Engagement: Highest monthly spend, buys strategically for resale
Categories
Underpriced cards, bulk lots, card breaks
Motivation
Arbitrage profit across platforms
Pain Points
Shipping costs, authentication concerns, tight margins

Seller Personas

The Side Hustler
$500-$5K/mo
Hours: 10-20 hrs/week, streams 2-3x, single niche
Background: Ex-eBay seller, flea market/thrift sourcer
"Sold in one stream what takes 3 months on eBay"
Pain Points
Shipping 20-50 packages, tax surprises, camera setup costs
The Full-Time Operator
$10K-$100K+/mo
Hours: 40-60 hrs/week, 1-3 employees, diversified categories
Scale: 500+ sellers doing $1M+ annually [↗], 62% exclusive to Whatnot [↗]
Pain Points
Sourcing inventory, 50-200 packages/day, chargebacks, Account Health anxiety
The Business/Brand
$50K-$500K+/mo
Setup: LLC with warehouse, professional production
Scale: Wholesale hit $1.5M/week after launch
Pain Points
Multi-platform coordination, inventory integration, sales tax compliance

Competitive Landscape

Payout timing, trust models, and strategic positioning

Payout Timing Comparison

Platform Standard Payout New Seller Hold During Dispute
eBay 1-2 days post-delivery Up to 21 days Held
StockX 2-4 days post-auth Same N/A
Mercari 3-day auto-release Same Held
Poshmark 3 days post-delivery Same Held
TikTok Shop 15 days post-delivery Same (15 days) Held
GOAT 2-3 days post-auth Same N/A
Amazon Biweekly (14-day cycle) 21-30 days Held

Note: Payout timings are based on platform seller documentation; verify before citing specific numbers in interview.

Trust Spectrum

Low Platform
Involvement
Craigslist Mercari eBay Poshmark Whatnot(?) GOAT StockX
High Platform
Involvement

Key Strategic Insights

eBay's Managed Payments: eBay's shift to Managed Payments was fundamentally a trust/fraud play, not revenue. Controlling money flow = dynamic holds based on real-time risk scoring.
Authentication Premium: StockX proves buyers accept 7-12 day delays if they trust authentication. For Whatnot collectibles, optional auth tier could command 15-30% premium.
Dynamic vs Blunt: TikTok Shop's 15-day hold is a blunt instrument. Dynamic risk-based payout timing is a competitive moat.

Trust & Risk Deep Dive

Fraud vectors, platform advantages, and partnership-first trust architecture

Whatnot's Current Policies

Counterfeit Items: Any product that copies or imitates a brand's trademarks, logos, or design without authorization. Includes items marketed as "replica," "inspired by," or "dupe" with brand-specific imagery.
"Unbranded Replicas" Loophole: Permitted outside Luxury Bags category. This creates a gray area where sellers can offer brand-inspired items as long as they don't explicitly use trademarks — except in luxury handbags where all replicas are banned.
Restricted Branded Items: Real Authentication required for luxury goods. Items over threshold values must pass through verification before payout.
Enforcement Gap: YouTube exposés (Spanish-language "RAD is RAD" channel) show scammers persist despite policies. Community-driven enforcement catches what automated systems miss.

6 Live Commerce Fraud Vectors

🃏
Card/Item Switching
Show premium item on camera, ship a fake or lower-grade substitute
📺
Stream Manipulation
Pre-recorded streams, deepfake overlays, misleading video quality
🤖
Shill Bidding
Fake accounts driving up prices to inflate final sale value
💰
Auction Manipulation
Cancel when bids too low, misleading start prices, phantom inventory
📦
Refund Fraud
Claim item not received, or send back a different/damaged item
🏦
Money Laundering
High-value self-buying to move money through the platform

4 Unique Advantages of Live Commerce

01
Video Record of Every Sale
Irrefutable evidence — every transaction is on camera with timestamps, creating an audit trail no other marketplace has.
02
Real-Time Community Policing
Chat flags suspicious behavior in real time. Viewers who know the category become the first line of defense.
03
Seller Identity is Visible
Harder to be anonymous when your face is on screen. Reputation is tied to a real person, not just a username.
04
Faster Feedback Loop
Issues surface in minutes, not days. A bad experience gets flagged before the stream ends.

Prevention ROI Framework

$1
Prevention
→
$10
Resolution
→
$100
Lost LTV

Every $1 in stream monitoring and seller vetting saves $10 in dispute resolution and $100 in lost buyer lifetime value.

Partnership-First Trust Stack

1
Pre-Transaction Trust
Persona (seller ID verification, $1-5/check) PSA/BGS (card auth, "Verified" badges) Entrupy (AI luxury auth during livestreams) CheckCheck (sneaker auth, ~$4/check)
2
Transaction Trust
Sift/Forter (ML fraud detection, serves Poshmark) Affirm/Afterpay (BNPL for $500+ items, +58% AOV) Escrow.com (ultra-high-value $5K+ protection)
3
Post-Transaction Trust
Route/AfterShip (shipping insurance, ~1.5% of value) AfterShip Tracking (branded tracking page) Chargebacks911 (prevention + representment with stream video)
4
Platform Trust
Verisart (blockchain provenance for high-value items) Market data APIs (Card Ladder, TCGPlayer price feeds) AI stream analysis (computer vision card pre-grading overlay)

Case Interview Toolkit

Framework, scenarios, metrics, and experience bridges

The 5-Step Framework

Step 1
Scope
→
Step 2
Map Journey
→
Step 3
Data
→
Step 4
Solutions
→
Step 5
Prioritize

5 Case Scenarios

Case A: Reducing Fraud/Counterfeits

Opening Question to Ask

"Are we seeing this more in specific categories, or is it platform-wide? And is the concern primarily buyer reports, or are we catching things proactively?"

Data to Request

Dispute rate by category, fraud reports per 1K transactions, chargeback rate trend (last 6 months), % of disputes from repeat offenders vs first-time, average time from purchase to fraud report.

Solution Layers

Quick win (rules): Mandatory seller ID verification for categories with >5% dispute rate. Hold payouts 48h for new sellers in high-risk categories.
Systemic (ML): Computer vision model trained on stream footage to flag item-switching (compare shown vs shipped). Behavioral scoring: sellers with unusually high cancellation + re-list patterns.
Platform design: "Verified Authentic" badge with optional third-party auth (PSA/BGS for cards). Price anomaly detection — items priced >50% below market get flagged.

Experience Bridge

BCP: Built layered fraud defenses for $600M lending portfolio — document fabrication, identity impersonation, embedded malware. Same principle: rules first, then ML, then systemic design.

Live Commerce Angle

The stream recording is an asset no other marketplace has. Every sale has video evidence. Build the dispute resolution around this — "show me the tape" should be the default.

Case B: Post-Purchase Buyer Experience

Opening Question to Ask

"What's our current post-purchase NPS, and where in the journey are we losing buyers? Is it between auction-close and delivery, or after delivery?"

Data to Request

Time from purchase to shipment (P50/P90), % of orders shipped within 48h, buyer re-purchase rate at 30/60/90 days, support ticket rate by issue type, delivery satisfaction score.

Solution Layers

Quick win: Automated shipping reminders to sellers at 24h/48h. Branded tracking page with stream replay link (AfterShip).
Systemic: Dynamic seller SLAs based on volume tier. Buyer "confidence meter" showing seller's on-time shipping rate before bidding.
Platform design: Post-delivery micro-survey (1 question: "Did you get what you expected?" with photo upload). Proactive outreach before buyer initiates dispute.

Experience Bridge

PayPal: 400M users, checkout team. Built AI system that cut issue detection from hours to minutes. Same principle: proactive beats reactive.

Live Commerce Angle

Link the stream clip to the order confirmation. Buyer can re-watch exactly what was shown. This builds confidence and reduces "that's not what I ordered" disputes.

Case C: Scaling Logistics

Opening Question to Ask

"Are we talking about scaling the existing seller-ships-direct model, or exploring Whatnot-facilitated fulfillment? And what's our current on-time delivery rate?"

Data to Request

Avg packages/day, on-time rate by carrier, shipping cost as % of GMV, international shipping failure rate, returns logistics cost, seller shipping compliance rate.

Solution Layers

Quick win: Negotiated carrier rates passed to sellers (volume leverage). Pre-printed shipping labels auto-generated after stream ends.
Systemic: Regional fulfillment hubs for top sellers (50-200 packages/day). Batch shipping optimization — sellers ship daily, not per-order.
Platform design: "Whatnot Shipping" program: sellers pay flat rate, Whatnot handles carrier selection + tracking. Start with top 100 sellers.

Experience Bridge

Trackly: Built classification engine processing 2,000+ companies. Same ops challenge: scaling a system that works at 100 to work at 10,000. Pattern recognition + automation.

Live Commerce Angle

Live commerce creates shipping spikes — a seller might sell 200 items in one 3-hour stream. Traditional e-commerce spreads orders. This needs stream-aware logistics planning.

Case D: Dispute Resolution System Design

Opening Question to Ask

"What's our current dispute resolution time, and what percentage get escalated beyond first-touch? Also, are disputes concentrated in specific categories?"

Data to Request

Disputes/1K transactions, median resolution time, % resolved at first touch, buyer satisfaction post-resolution, seller satisfaction post-resolution, cost per dispute.

Solution Layers

Quick win: Auto-resolve disputes under $25 in buyer's favor (cost of resolution > item value). Stream recording auto-attached to every dispute.
Systemic: Tiered dispute system: auto-resolve → community moderator → Whatnot specialist → expert panel. ML-powered triage based on dispute type + value + seller history.
Platform design: "Dispute Prevention Score" for sellers. High-risk sellers get mandatory photo verification at packing. Transparent resolution timelines shown to both parties.

Experience Bridge

BCP: Risk-scored 100K small business entrepreneurs. Dispute resolution is fundamentally a risk-scoring problem — who's more likely telling the truth, given their history?

Live Commerce Angle

The video record is transformative for disputes. Build a tool where the dispute reviewer can jump to the exact moment the item was shown in the stream. Compare to what was shipped.

Case E: Seller Churn from Bad Buyers

Opening Question to Ask

"Is this concentrated among high-value sellers, or broad-based? And what are the top 3 buyer behaviors sellers cite when they leave?"

Data to Request

Seller churn rate by revenue tier, top reasons cited in exit surveys, refund abuse rate (buyers with >3 refunds/month), non-payment rate, seller NPS trend.

Solution Layers

Quick win: Buyer reliability score visible to sellers (on-time payment %, dispute rate). Auto-block buyers with >15% refund rate from bidding.
Systemic: Buyer Account Health — mirror the seller health concept. Restrict bidding privileges for serial abusers. "Trusted Buyer" badge for top 10%.
Platform design: Seller protection fund — Whatnot covers losses from verified buyer fraud. Revenue share on prevented chargebacks. Seller success manager for top 500 sellers.

Experience Bridge

Trackly: Built classification system that learns from human feedback. Same principle: use seller reports to train buyer risk models. Feedback loops improve over time.

Live Commerce Angle

Sellers can see who's in their audience. Give sellers tools to manage their stream — quiet mode for known trolls, priority bidding for trusted buyers.

Metrics Reference

Trust & Safety
  • Dispute rate per 1K transactions
  • Chargeback rate (% of GMV)
  • Fraud detection precision/recall
  • Counterfeit interception rate
  • Time to resolve disputes (P50/P90)
  • Repeat offender rate
Buyer Experience
  • Post-purchase NPS
  • Re-purchase rate (30/60/90 day)
  • Delivery satisfaction score
  • "Item as described" rate
  • Support tickets per 1K orders
  • Time from bid to delivery (P50)
Seller Health
  • Seller churn rate (monthly)
  • Ship-on-time rate
  • Seller NPS
  • Avg payout timing
  • Listings per active seller
  • Revenue per seller (by tier)
Operational
  • Cost per dispute resolution
  • CX tickets per 1K orders
  • Auto-resolution rate
  • Shipping cost as % of GMV
  • On-time delivery rate
  • International shipping failure rate

Experience Bridges

BCP → Whatnot
Risk-scoring 100K entrepreneurs → Risk-scoring sellers/buyers
Document fabrication detection → Counterfeit detection
Layered fraud defenses (rules + ML + design) → Trust stack architecture
$600M revenue protected → GMV protection at scale
PayPal → Whatnot
400M user checkout → Platform-scale ops
AI issue detection (hours to minutes) → Real-time fraud detection
15 distributed systems monitoring → Multi-service marketplace orchestration
Payment flow expertise → Payout timing optimization
Trackly → Whatnot
Classification engine (2K companies) → Item categorization at scale
Rules system with learning loops → Adaptive trust rules
Full-stack builder (DB to app) → Cross-functional ops execution
Monitoring 299 Whatnot jobs → Deep operational understanding

Tactical Playbook

Scripts, questions, pitfalls, and power moves

TMAY (Tell Me About Yourself)

~75 sec
I'm Kevin, a PM with 7 years of experience, mostly in fintech. Three things relevant to this role:

First, I know risk and trust systems at scale. At BCP, Peru's largest bank, I risk-scored 100K small business entrepreneurs for lending products. I dealt with three types of fraud — document fabrication, identity impersonation, and embedded malware — and built layered defenses for each. That's $600M in annual revenue protected.

Second, I've operated at platform scale. At PayPal, I was on the checkout team serving 400 million users. I built an AI system that cut issue detection from hours to minutes across 15 distributed systems.

Third, I build. I shipped Trackly, an AI recruiting agent that monitors 2,000 companies. I built the classification engine, the rules system — everything from the database to the app.

I'm excited about Whatnot because I actually bought on the platform, and I've been thinking about the post-purchase experience.

Why Whatnot

~45 sec
Three reasons.

First, the problem is real. Live commerce is genuinely different from traditional e-commerce — the trust dynamics, the real-time nature, the community element. These aren't problems you can copy-paste solutions from eBay to solve. That excites me.

Second, the stage. $8B GMV [↗], scaling internationally, 250+ categories — this is the messy, high-stakes operational scaling where my experience at BCP and PayPal is most relevant.

Third, I actually use the product. I've bid on cards, watched streams, experienced the post-purchase flow. When I think about improvements, it's from a user perspective, not just theory.

Questions to Ask

During Intro (pick 2)

At End (pick 3-4)

8 Pitfalls to Avoid

01
Jumping straight to a solution without asking questions
Ask 2-3 scoping questions first. "Before I propose anything, can I understand..."
02
Leading with "we should build an ML model"
Start with rules/heuristics, then layer complexity. "First, simple rules. Then, as we have data..."
03
Monologuing for 3+ minutes without checking in
Pause every 60-90 seconds. "Should I go deeper on this, or pivot to another angle?"
04
Comparing unfavorably to TikTok Shop or eBay
Frame competitors as reference points, not as better. "eBay solved X this way, but Whatnot has Y advantage"
05
Forgetting the seller side of the marketplace
Always address both sides. "This helps buyers because X, and sellers because Y"
06
Presenting solutions without tradeoffs
Name the downside. "The tradeoff here is we'd slow down seller onboarding, which means..."
07
Being too academic or framework-heavy
Use frameworks silently. Structure your thinking, but talk in concrete terms.
08
Saying you bought on the app more than once
Mention it once, early, brief. Then let your knowledge speak for itself.

6 "Impress" Moves

01
Name specific data before proposing a solution. "I'd want to see dispute rate segmented by category and seller tenure before deciding."
02
Reference the video record advantage. "Live commerce gives us something eBay doesn't — every sale is on tape."
03
Think in layers: quick win (this week) + systemic solution (this quarter) + platform moat (this year).
04
Quantify impact. "If dispute rate is 5% and we cut it to 3%, on $8B GMV [↗] that's $160M in protected transactions."
05
Show marketplace empathy. "We need to be careful not to over-burden sellers with verification, or we lose the 62% who are exclusive to Whatnot." [↗]
06
Connect to your experience naturally. "At BCP, we faced a similar cold-start problem with new borrowers and solved it with..."

Cheat Sheet

Screenshot this. Review 5 minutes before the call.

LIVE CASE CHEAT SHEET

  • Ask 2-3 questions BEFORE proposing solutions
  • Start simple (rules), then layer complexity (ML/systems)
  • Name the data you'd look at before the solution
  • Quick win + systemic solution for everything
  • Pause every 60-90 seconds — "Should I go deeper?"
  • Tradeoffs: every solution has a downside, name it
  • Bridge to: BCP fraud ($600M), PayPal scale (400M users), Trackly classification (2K companies)
  • Mention you bought on the app (once, early, brief)
  • NEVER: Monologue for 3+ minutes without a pause
  • NEVER: Compare unfavorably to TikTok Shop
  • NEVER: Jump to ML first — always start with rules
  • NEVER: Forget the seller side of the marketplace

Practice Drill

Run through this once before the interview. Time yourself.

"We've been seeing an increase in counterfeit trading cards. Walk me through how you'd approach this."
0:00 - 0:30
Scope Question
"Before I dive in — is this primarily in graded cards, raw cards, or both? And are we seeing it more from new sellers or established ones?"
0:30 - 1:00
Data Request
"I'd want to look at: dispute rate for trading cards vs platform average, % of counterfeits from repeat offenders, and whether there's a price threshold where fakes concentrate."
1:00 - 1:45
Quick Win
"Immediate: require photo of card front + back before shipping for items over $50. Cross-reference with the stream recording. Flag mismatches automatically."
PAUSE
Check In
"Does this direction make sense? Should I go deeper on the quick win, or move to the systemic solution?"
1:45 - 2:45
Systemic Solution
"Longer term: partner with PSA/BGS for a 'Verified Authentic' badge program. Sellers who opt in get faster payouts and higher visibility. Build a computer vision model trained on stream footage to detect card-switching — compare what was shown on camera to what was photographed for shipping."
2:45 - 3:15
Live Commerce Angle
"What's unique here is the video record. Every sale is on tape. We can build dispute resolution around 'show me the tape' — this is an advantage eBay and StockX don't have."
3:15 - 3:30
Tradeoff
"The tradeoff: requiring photo verification adds friction to shipping. We'd need to make it as seamless as possible — maybe in-app camera with auto-crop — or we risk slowing down our fastest sellers."
PAUSE
Let Interviewer React
"Those are my initial thoughts. What aspect would you like me to go deeper on?"

Total: ~3.5 min talking • 2 pause points

This is a conversation, not a presentation. The pauses are where they steer you.

Sources & Verification

Every major data point with its primary source and verification status

Verified — Direct primary source available

Data PointSourceStatus
$8B GMV (2025)Sacra — sacra.com/c/whatnot/Verified
$11.5B valuation, Series F Oct 2025CB Insights — cbinsights.com/company/whatnotVerified
$225M Series FCB Insights + SacraVerified
Series E: $265M at $4.97B (Jan 2025)TechCrunchVerified
20M+ new accounts (2025)SacraVerified
#1 Shopping app US & UKSacra (originally from Whatnot press)Verified
~60% market share (NA + Europe)SacraVerified
1.5M+ daily transactionsSacraVerified
~95 min/day engagementSacra (range: 80-95 min)Verified
80%+ MoM retentionSacraVerified
Gen Z: 16 orders/yearSacraVerified
374% increase first-time buyersSacraVerified
Beauty +791%, Electronics +444%, Jewelry +259%, Fashion +223%SacraVerified
35+ new categories in 2025SacraVerified
Wholesale $1.5M/weekSacraVerified
1 in 8 sellers full-time (up 20% YoY)SacraVerified
500+ sellers at $1M+SacraVerified
62% seller exclusivitySacraVerified
European sellers grew 600% YoYSacraVerified
Australia: hundreds of shows $10K+SacraVerified
Pre-bids: 7x more salesSacraVerified
$100M+ Black Friday (48h)Sacra (note: conflicting $75M figure also exists)Verified
299 jobs tracked, hiring distributionTrackly LAMP Monitor (Kevin's own database)Verified

Estimated — Derived from verified data or single third-party estimate

Data PointSourceStatus
~$1B revenueSacra estimate based on 12.5% take rate on $8B GMV (not officially disclosed)Estimated
~1,745 employeesLinkedIn — linkedin.com/company/whatnot-inc/Estimated
Live commerce conversion ~30%Widely attributed to McKinsey "It's Showtime" (2021); exact figure unverifiableEstimated
US live commerce market ~$15BMultiple research firms; exact figures behind paywallsEstimated

Unverified — Could not find primary source; use with caution

Data PointSourceStatus
MrBeast 500K+ concurrent viewersWhatnot engineering blog describes the event but exact viewer count unverified from accessible sourcesUnverified
France specific growth %Conflicting figures (888% vs 427%); neither independently verified. Removed from prep.Unverified
500K hours of live shows/weekOnly 20K hours/week confirmed for Europe (Sacra). Global figure not verified.Unverified
Sellers streaming 3-4x/week avg $13K+/monthLikely from Whatnot's "State of Live Selling" report (not publicly accessible)Unverified
52% plan to hire, 88% live selling beliefSame source; not publicly accessibleUnverified
TikTok Shop 15-day hold, eBay 21-day holdDirectionally plausible; seller docs not fetched for verificationUnverified
China live commerce ~$800BRange estimates vary ($500-800B); no single verified sourceUnverified
50+ categories planned for 2026Forward-looking; no source foundUnverified
Cross-category buying +75%Not found in any accessible sourceUnverified
12M+ monthly fashion ordersNot found in any accessible sourceUnverified

Prepared with Trackly LAMP data (299 jobs analyzed) + Sacra + CB Insights + TechCrunch + Whatnot Engineering Blog + Trustpilot • All data points source-linked — see Sources section • Last updated: April 12, 2026