question bank
Every question, answered against the bar.
Not a list of prompts. Each question shows a strong answer next to a weak one, the framework that fits, and who asks it.
01 Design an alarm clock for people who are blind.How to answer this Microsoft PM classic without the braille-and-voice checklist. The deaf-blind variant shows up in senior loops and most candidates fail it. standard → 02 How would you design a bicycle renting app?A product sense question on dockless bike-sharing. What separates a strong answer from a generic one is supply-side thinking and unit economics. standard → 03 Design a communication tool for children.A Stripe, Google, and Meta product design question. How to name the dual-user tension, pick a real segment, and clear the bar in 2026. hard → 04 Design a product that connects handymen with consumers.A Meta product sense staple. The strong answer solves both sides of the marketplace and names the cold-start problem without being prompted. standard → 05 Design a new feature for Discord.The Discord product sense question. How to pick a specific persona, name the real pain, and weave safety in from the start instead of tacking it on. standard → 06 Design a jobs product for Facebook.How to answer this Meta product-sense question with the social graph mechanic, the right user segment, 2025 relaunch context, and metrics that clear the bar. hard → 07 Design a news feed.The complete PM interview answer for designing a news feed in 2026: objective function choice, AI ranking, creator economics, and EU regulatory constraints. hard → 08 Design a product for commuters.How to give a senior-level answer to this Lyft, Uber, and Google PM question in 2026, when real-time tracking is already a commodity. standard → 09 Design a product for dog owners.How to answer this PM design question with a real user segment, a non-obvious pain, and a 2026 lens on viability, lovability, and competitive moat. standard → 10 Design a product for the elderly.How to pick the right segment, name the real job, address the buyer/user split, and avoid the generic accessibility trap on this PM product-sense question. standard → 11 Design a product for new parents.How to answer this PM product-sense question with a sharp segment, fourth-trimester data, and a concept that clears the viable and lovable bar. standard → 12 Design a product for pet owners.How to answer this PM product-sense question by scoping a vague user group, picking a segment with proof, and passing the 2026 viable and lovable bar. standard → 13 Design a product for the elderly.Model PM answer for "design a product for the elderly." Segment selection, real JTBD, metrics with targets, and the 2026 data that kills the weak answer. standard → 14 Design a product for college students.How to answer this PM interview prompt in 2026. Pick a sub-segment, surface the institutional buyer, and propose something lovable, not just usable. standard → 15 Design a product for people who are blind or have low vision.How to answer the "design a product for the blind" PM interview question without designing from assumption or reinventing tools that already exist. standard → 16 Design a product for volunteers.A complete worked answer for the Meta PM product-sense prompt. Covers the real gap in the market, who pays, and what separates a pass from a fail. standard → 17 Design a refrigerator for the blind.A worked PM interview answer with the right framework, a strong and weak example, and why 2026 AI changes the correct answer entirely. standard → 18 How would you design a scooter sharing app?A product-sense interview question. The strong answer addresses viability tension, unit economics, and the commuter habit loop, not just map and QR unlock. standard → 19 Design a travel app for Google.How to answer Google's travel app product sense question: what interviewers score, what strong vs. weak looks like, and what Google-specific thinking means. hard → 20 How would you design WhatsApp for students?A complete PM interview answer for the WhatsApp student design question: segment, friction, features, legal constraints, metrics, and the 2026 AI angle. hard → 21 What is your favorite product and why?How to answer the favorite-product PM interview question with product judgment instead of fan enthusiasm. Updated for 2026. warmup → 22 Only about 10% of Gemini tutor (Guided Learning) users describe their experience as "magical." Why, and what's the highest-leverage thing you'd fix?Google/DeepMind product sense and execution hybrid. Diagnose why Guided Learning misses with 90% of users, pick one fix, define magical with real metrics. hard → 23 How would you improve Gmail?A complete model answer for the Gmail product-sense question in 2026, including what strong and weak answers look like and where candidates typically fail. standard → 24 How would you improve Facebook Groups?A complete Meta product sense answer for Facebook Groups in 2026, with the right user cut, a focused solution, and the strategic tension that clears the bar. standard → 25 How would you improve Google Maps to increase SMB adoption?A two-sided marketplace answer for Google Maps SMB adoption interviews. Covers supply-side activation, SMB segments, 2026 AI stakes, and real metrics. hard → 26 How would you improve Google Maps?The 2026 answer pivots from navigation to trust. Here is what separates a strong answer from a recycled framework on this over-asked question. standard → 27 How would you improve Google Maps?Pick a user, find a real unmet need, defend one bet. How to clear the product-sense bar at Google in 2026 without proposing features that already launched. standard → 28 How would you improve Instagram Reels?A model answer for Meta product-sense interviews. Scope to mid-tier creators, address the viability gap, and name the intra-product cannibalization tension. standard → 29 How would you improve LinkedIn engagement?A rigorous sample answer to LinkedIn's classic product-sense question, covering the creator supply problem, identity risk, and the right north star metric. hard → 30 How would you improve the Uber app?A step-by-step answer to the Uber product-sense question, covering AV markets, Uber One, and what separates strong answers from rehearsed frameworks. standard → 31 How would you improve Spotify's engagement?A model answer for the Spotify product-sense interview. Scope the metric, name the real failure modes, and build on Spotify's 2026 strategic bets. hard → 32 How would you improve WhatsApp?The Meta product sense question candidates get wrong by pitching features that exist. A model answer built around business messaging, not student use cases. standard → 33 How would you improve YouTube's recommendations?A model answer for Google/YouTube PM interviews. Scope the surface, reject watch time as North Star, and defend satisfaction-weighted discovery. standard → 34 What is your least favorite product and how would you fix it?The least-favorite product question is a product-improvement exercise in disguise. Here is the answer structure, a worked example, and the rubric signal. standard → 35 Imagine you're a PM for Meta. Design a product for Volunteering. What would you build? Why?How to pass Meta's 60-minute Product Sense with AI interview using the volunteering app prompt, covering both halves and the human delta test. hard → 36 Solve the dog poop problem.A product-sense question that tests viability, not UX. Who pays, does the economics work, and why does the $250 NYC fine produce 2 tickets a year? standard → 37 Technology now lets humans understand animals. What would you build?A real OpenAI PM interview question. How to clarify the tech constraint, pick a viable segment, and scope an MVP that passes the viability test. hard → 38 What makes a product well-designed?A two-layer framework for product sense interviews: usability is the floor, lovable is the bar. Includes strong and weak sample answers. standard →
01 How would you define the north star metric for this feature?How to pick a north star metric for a specific feature in a PM interview, including the AI feature variant and why DAU always fails the test. standard → 02 How would you design an A/B test for a core product flow?How to structure an A/B test answer in a PM interview: hypothesis, metrics, guardrails, and the 2026 wrinkle of non-deterministic AI features. standard → 03 How would you measure the success of Facebook Marketplace?Framework for Marketplace metrics at the Meta bar: north star, two-sided dynamics, trust guardrails, and the 2026 AI feature layer. standard → 04 How would you measure the success of Google Maps for small businesses?A Google PM interview question on measuring SMB engagement in Maps. Covers the two-sided market, North Star selection, and the 2026 AI Overviews shift. hard → 05 How would you measure success for Uber Eats?The three-sided marketplace metric question. Name a north star, defend it against alternatives, and map the decision vs. diagnostic split interviewers test for. hard → 06 How do you decide between a phased rollout and a full launch?A decision framework for the PM interview question on phased vs full launch, covering reversibility, blast radius, kill switches, and AI-specific rollout risk. standard → 07 When is an MVP ready to ship?A three-gate framework for PM interviews on MVP readiness, including what interviewers penalize and the 2026 AI shift. standard →
01 Daily active users dropped 20% week over week. Find the root cause.A diagnostic sequence for the PM RCA interview. Covers the measurement-error trap, step-change vs. trend bisect, and AI-product failure modes. hard → 02 Facebook Groups engagement dropped 15% over the past month. How do you find the root cause?A Meta RCA interview question. Covers the measurement-error check, segmentation tree, AI feed ranking hypotheses, and how to commit to a cause. hard → 03 You're a PM at Instagram. Feed post volume has declined 15% over the past six months while Stories usage has grown. Is Stories cannibalizing posts? What do you do?How to diagnose and respond when Stories growth correlates with a feed post drop. RCA, cannibalization vs. substitution, and what to do next. hard → 04 Notification engagement is up six weeks running, but time on site is flat. What do you do?A worked answer for the Meta PM interview question where notification engagement rises but time-on-site stays flat. Covers the full hypothesis tree. hard → 05 Facebook Groups engagement dropped 15%. Diagnose the root cause.How to structure a Meta RCA interview answer on Groups engagement: measurement check, supply vs demand split, and the 2026 AI-feed hypothesis. hard → 06 Instagram MAU is flat but DAU has been declining for the past several weeks. Walk me through how you would diagnose this.How to diagnose a metric divergence in a Meta PM interview. MAU flat + DAU down is a deductive constraint, not just context. hard → 07 Messenger DAU dropped 8% last week. Walk me through how you'd find the root cause.How to diagnose a Messenger DAU drop in a Meta PM interview, from data integrity checks to cannibalization hypotheses and the 'so what' close. hard → 08 There is a data point showing more Uber drop-offs at the airport than pickups. Why might that be, and what would you do about it?Uber airport imbalance RCA: supply-side driver economics, FIFO queue mechanics, and what separates a strong answer from a pattern-match. standard → 09 Reddit traffic is down 5% week-over-week. What do you do, and what do you tell the exec team?A complete RCA answer for the Reddit traffic drop interview question, covering the diagnostic process and exec communication under uncertainty. hard → 10 Retention dropped 5% last week. Walk me through how you'd diagnose it.How to run a retention RCA in a PM interview. Covers data integrity gates, cohort decay types, segmentation order, and 2026 AI-cause additions. standard → 11 A key streaming metric dropped 80% overnight. Walk me through your root cause analysis.A streaming-specific RCA walkthrough for PM interviews. Covers the 80% magnitude signal, hypothesis tree, and 2026 AI model failure modes. hard →
01 How would you design an A/B test for an AI feature whose outputs are non-deterministic?The AI PM experiment design question. How to control model, evaluation, and population variance when your feature's output changes every run. hard → 02 Design the guardrails for an agent that can act on a user's behalf across their calendar, inbox, and task management tools.The 4-dimension rubric for agent guardrails interview questions: scope, confirmation, limits, and kill switch. What clears the bar at AI-first companies. hard → 03 Design an evaluation framework for a customer-support AI agent before it talks to real customers.How to answer the AI PM eval-design question. Offline vs online evals, real thresholds, MiHR/MaHR, shadow mode, and the flywheel that makes evals a product moat hard → 04 How would you make an LLM's output more creative?The AI PM creativity question. Three control layers, the sampling-and-ranking pattern, eval signals, and what clears the bar at Google and Apple. standard → 05 How would you define goals and measure success for a new AI product?The AI metrics interview question that separates candidates who've shipped from those who've read about it. A step-by-step framework with a worked example. hard → 06 How would you optimize an LLM for retrieval on a given dataset?The Meta AI PM retrieval question decoded. Decision hierarchy, RAG vs fine-tuning, hybrid search, eval metrics, and the 2026 reframe interviewers are testing for. hard → 07 How would you price an AI product without killing margin?The AI PM pricing question. Token economics, the autonomy-attribution matrix, and margin levers, structured as a spoken interview answer. hard → 08 When would you use prompting, RAG, or fine-tuning for an AI feature?The AI PM decision question. A decision framework, 2026 production realities, follow-up prep, and what clears the bar at OpenAI and Anthropic. standard → 09 What safeguards would you build into an agentic AI product, and how would you decide which ones to prioritize?How to answer the agentic AI safeguards question in PM interviews. The irreversibility framework, key vocabulary, and what separates a hire from a pass. hard → 10 What hallucination rate would cause you to refuse to launch?How to answer the AI PM launch-gate question with a specific threshold, measurement method, and fallback path that clears the bar. hard → 11 How would you validate an AI model before shipping it?What a strong PM answer looks like in 2026: offline gates, online monitoring, grader design, agentic evals, and the viability/lovability layer beyond hard → 12 When does a feature warrant an agent rather than a simpler LLM call?The three-gate framework for deciding when to use an AI agent, with strong/weak answer examples and the 2026 feasibility reframe. hard →
01 Why did Amazon enter the hardware business?The Amazon strategy question every PM candidate should own. Three nested motivations, the ad revenue flywheel, and the 2026 AI-agent layer explained. hard → 02 What is the biggest threat to Reddit?A strategy answer calibrated to the strong-hire bar. The real threat is not TikTok: it is AI content eroding Reddit's authenticity signal from within. hard → 03 How do you decide whether to build a capability in-house, buy a vendor solution, or form a partnership?The PM strategy interview question on build vs buy vs partner. A decision framework built around differentiation, data compounding, and reversibility. hard → 04 Should Meta consolidate Messenger, Instagram DMs, and WhatsApp into a single messaging product?A Meta strategy question on messaging unification. Separate the three consolidation layers, give a clear recommendation, and handle the DMA constraint. hard → 05 You're the CPO of Zoom. How do you compete with Microsoft Teams and Google Meet over the next two years?A CPO-level strategy question on Zoom vs Teams. Model answer, weak answer, and the 2026 competitive facts you need to sound credible. hard → 06 How should this company monetize its product?A PM interview framework for monetization strategy questions. Decision logic for choosing ads, subscriptions, usage-based, or outcome-based models. hard → 07 How would you price a new product and build its go-to-market strategy?Price and GTM are one question. This covers the sequencing, the strong answer, the weak answer, and the 2026 AI viability context. hard → 08 A competitor just launched your feature. What do you do?How to answer the competitive response PM interview question. Diagnosis before action, the urgency trap, and what separates a strong answer from a panicked one. hard → 09 Should Amazon enter the e-learning market?A model strategy answer for PM interviews. Which segment, why B2B AI upskilling, what Amazon already has, and the 2026 context that changes the frame. hard → 10 Should Amazon enter the food delivery market?A strategy answer grounded in what Amazon actually did. Build vs. buy vs. partner, the Grubhub facts, and the kill criterion that clears the bar. hard → 11 Should Google build a streaming service?A Google strategy interview question on build vs. partner, content economics, and where Google's real moat sits in the streaming value chain. hard → 12 Should Meta build a dating product?A model answer for the Meta PM strategy question. Recommendation, evidence, North Star, and 2026 AI framing. Not a framework recitation. hard → 13 What should Meta's next acquisition be?A model answer for the Meta M&A strategy interview question. Gap logic, a named 2026 target, and why TikTok is the wrong answer. hard → 14 Should Meta enter the education market?A strategy answer that treats Meta's legal exposure with minors as a structural constraint, not a risk to mitigate. The right entry point is not K-12. hard → 15 Design a go-to-market strategy for a text-to-music AI product.How to answer the text-to-music GTM question in a PM interview, with real market data, a worked strong answer, and the traps to avoid. hard →
01 How many e-scooter rides happen in San Francisco in a month?A complete worked estimation answer for the SF e-scooter rides PM interview question, with real SFMTA data, dual approach, and sanity check. standard → 02 Estimate the size of the US electric vehicle market.A worked bottom-up estimation for PM interviews, with 2025-2026 data anchors, three valid approaches, and the policy shock most candidates miss. standard → 03 Estimate the number of gas stations in the United States.A worked PM estimation answer with two independent approaches, a sensitivity driver step, and the 2026 EV follow-up. Structure beats arithmetic. warmup → 04 How many queries does Google handle per second?A worked bottom-up estimation of Google QPS for PM interviews, including scope clarification, user segmentation, and the 2026 AI Overview cost layer. standard → 05 How many iPads are sold in the US per year?A dual-method estimation walkthrough for PM interviews. Bottom-up household model plus a top-down Apple revenue cross-check, with the business insight layer. standard → 06 How would you estimate the size of the US pet food market?A worked bottom-up estimation of the US pet food market with real anchor numbers, a sanity check, and the PM "so what" that separates a hire from a pass. standard → 07 How many messages does Slack send per day?A step-by-step estimation for PM interviews. Segment DAU by user type, layer in bot traffic, and sanity-check against the 1.5B anchor. standard → 08 How many piano tuners are there in New York City?Worked Fermi estimation for PM interviews. Full decomposition, institutional segment, sanity check, and what the interviewer actually scores. warmup → 09 How many restaurants are in Mumbai?A worked estimation for the PM interview question on Mumbai restaurants, covering organised vs. unorganised sectors, cloud kitchens, and the 2026 product angle. standard → 10 How would you estimate the annual revenue of a single Times Square billboard?A step-by-step estimation framework for Times Square billboard revenue, with real CPM, rotation, and occupancy numbers you can defend out loud. standard → 11 How many Uber rides happen in New York City per day?A worked PM estimation with demand-side and supply-side methods, real TLC anchor data, and NYC-specific constraints. Builds to ~550K with a range of 500-600K. standard → 12 How would you estimate the size of the US toilet paper market?A worked bottom-up estimation of the US toilet paper market with consumer vs. institutional split, real anchor numbers, and the PM so-what. standard → 13 How many windows are in New York City?A PM estimation with borough-by-borough building-type segmentation, real NYC housing data, the curtain-wall architectural nuance, and the 2026 interviewer expectations. standard →
01 What is the difference between a leading and lagging indicator? Give an example.What interviewers actually test, why NPS is the wrong example, and how to answer with Slack and Facebook as evidence. standard → 02 How would you measure the success of Instagram Stories?How to pick a defensible north star for Instagram Stories, reject the wrong candidates, and handle cannibalization and Reels follow-ups at Meta. standard → 03 How would you measure the success of Google Photos?A defensible north star for Google Photos in 2026, with a full metric tree, counter-metrics, and the Google One monetization angle most candidates miss. standard → 04 What are Netflix's top 3 metrics?The right answer reflects Netflix's 2025 shift away from subscriber counts. Know the three metrics they actually report and why the hierarchy matters. standard → 05 How would you measure the success of WhatsApp?North star, retention, network depth, and the Business API monetization bridge. A structured answer for Meta PM interviews. standard → 06 Engagement is rising but revenue is falling. What do you do?The two-metrics conflict question. Four root causes, a worked diagnostic, the weak answer to avoid, and the 2026 AI angle. hard → 07 Which metric would you never optimize?The counter-intuitive PM interview question. How to name a specific metric, explain the Goodhart's Law trap, and hold your position under follow-up. standard →
01 Tell me about a time an AI product you worked on failed publicly. How did you respond?How to answer the AI failure recovery interview question, including failure type taxonomy, rollback decisions, and trust-signal redesign. hard → 02 Tell me about a time you demonstrated [Leadership Principle].All 16 LPs, Bar Raiser probe logic, L5 vs L6 calibration, a story-bank strategy, and the 2026 AI-era reframe for PM candidates. hard → 03 Tell me about a time you had a conflict with an engineer. How did you handle it?What interviewers actually check for in the PM-engineer conflict question, with a strong answer, a weak answer, and level-specific calibration. standard → 04 Tell me about a time you received difficult feedback. How did you handle it?What interviewers actually score, why most answers fail, and how to calibrate for mid-level vs. senior at Amazon, Google, and AI-native labs. standard → 05 Tell me about a time you disagreed with your manager.What the interviewer is actually scoring, how the bar shifts at senior level, and a concrete PM-specific answer you can adapt to your own story. standard → 06 Tell me about a time you drove a project end-to-end with incomplete information.What interviewers actually grade when they ask this behavioral question, plus a full strong answer, a full weak answer, and the 2026 angle. standard → 07 What is your greatest weakness?How to answer the greatest weakness question in PM interviews without sounding rehearsed or getting screened out. standard → 08 Tell me about a time you influenced without authority.What interviewers actually score on this question, why common answers fail, and how to structure a strong answer at senior PM level. standard → 09 Tell me about a project you killed that you loved.What this behavioral question actually tests, why most candidates pick the wrong story, and how to structure an answer that clears the bar. hard → 10 Tell me about a time you said no to a stakeholder.What this behavioral question actually tests, a full PM-specific example answer, the follow-ups that break rehearsed stories, and the 2026 viability reframe. standard → 11 Tell me about a time you failed.How PMs answer the behavioral failure question: story selection, SPSIL structure, time allocation, and the 2026 AI PM angle. standard → 12 Tell me about a time you used data to influence a decision.How to answer the PM behavioral question that scores data fluency and stakeholder influence simultaneously, including the 2026 AI PM layer. standard → 13 Why do you want to be a product manager?What interviewers actually score on, the exact failure modes, and a 2026-aware model answer for career-switchers and first-timers alike. warmup → 14 Why should we hire you?How to answer "why should we hire you" in a PM final round. Structure, a strong example, a weak example, and the 2026 judgment framing. standard → 15 Why do you want to work here?What interviewers actually score on this question, the failure modes that end candidacies, and a worked example answer you can adapt. warmup →
01 How do you prioritize your backlog?A strong answer names what you are not building and why. This tests judgment, confidence scoring, and 2026 viability thinking, not framework recall. standard → 02 Two executives each want you to prioritize their feature. How do you decide?A prioritization question that tests political navigation, evidence use, and conviction under pressure. Not framework recall. hard → 03 You are the PM for Facebook Live. What features do you prioritize?How to prioritize Facebook Live features in a Meta PM interview, including the ghost-stream problem, north star metric, and what clears the bar. hard →
01 How would you handle API versioning as a PM?What interviewers test when they ask about API versioning, plus a strong answer with Stripe and Twilio specifics and the 2026 agentic angle. standard → 02 How would you design the elevator algorithm for a 40-story office building with an average of 100 people per floor during a standard 9-to-5 workday?A verified Google PM technical interview question. Know the four algorithms, the multi-elevator dispatch problem, and the 2026 ML dispatch angle. standard → 03 Explain caching to a non-technical executive, and tell me the tradeoff you'd own as the PM.The PM technical round tests communication and judgment, not implementation depth. Here is what separates a 3/5 from a 5/5 answer. standard → 04 How would you explain embeddings to a non-technical executive?How to explain embeddings in a PM interview without losing an exec or sounding like you memorized the docs. Includes strong/weak answers and follow-up traps. standard → 05 Can you explain GET, POST, PUT, and DELETE, and what does idempotency mean?How to explain HTTP methods and idempotency in a PM interview. Safe vs idempotent, PATCH clarified, and why retry logic is a product decision. standard → 06 Can you explain the tradeoffs between REST and GraphQL to a non-technical executive?What interviewers are actually testing, a model exec-ready answer, and the 2026 hybrid-first reality most candidates miss. standard → 07 What happens when you type a URL into your browser and press Enter?The PM technical interview question. Walk the DNS-to-render chain at the right depth, name the PM-relevant trade-offs, and explain it to a non-technical exec. standard → 08 What is the difference between a database and a cache?The PM technical literacy question. How to explain database vs cache in 90 seconds, with the tradeoff framing interviewers actually want to hear. warmup → 09 What is an API, and what would you consider before building one?How to answer "what is an API" in a PM interview. Plain-English definition, PM considerations, and what Stripe, Twilio, and OpenAI reveal. standard →
01 Design Instagram's system.How to answer the Instagram system design question as a PM: social graph tradeoffs, hybrid fan-out, 2026 recommendation architecture, viable/lovable framing. hard → 02 Design Netflix's system.How PMs answer "design Netflix's system" at the right depth: Open Connect, recommendation architecture, and the viable/lovable lens for 2026. hard → 03 Design a notification system.How to answer the notification system design question as a PM, covering channel tradeoffs, frequency controls, and the 2026 AI-agent angle. hard → 04 Design a rate limiter.How to answer "design a rate limiter" as a PM. Covers algorithm tradeoffs, distributed state, 429 UX, pricing tiers, and 2026 AI token-budget problems. hard → 05 Design TikTok's system.How to answer the TikTok system design question as a PM: FYP pipeline, cold-start tradeoffs, 2026 signal hierarchy, and the creator-consumer flywheel. hard → 06 Design Twitter's tweet feed and timeline system.What PMs are evaluated on in the Twitter/X system design interview. Trade-offs, the celebrity problem, and the For You vs. Following tension. hard → 07 Design Uber's ride-matching system.What PMs are actually evaluated on in the Uber system design interview. Dispatch, surge, marketplace dynamics, and the 2026 AI layer. hard → 08 Design a URL shortener.How to answer "design a URL shortener" as a PM, not an engineer. Covers user types, the analytics business model, 301 vs 302, and 2026 abuse risks. standard → 09 Design WhatsApp.What PM interviewers actually score in this question, covering E2E encryption tradeoffs, delivery guarantees, and the AI-in-chat tension. hard → 10 Design YouTube.How PMs answer "design YouTube" at the right altitude: clarifying questions, CDN trade-offs, recommendation architecture, and the creator-viewer tension. hard → 11 Walk me through a system you designed. What were the tradeoffs?What Stripe actually evaluates in this round, how to pick the right system, and what separates a pass from a fail on failure-mode reasoning. hard →