Your site's "face," scored against the best

HeroHeadline: "The product development system for teams and agents" — Sub: "Purpose-built for planning and building products. Designed for the AI era."

How-it-works deviceNAMED-STAGE SCROLLYTELLING SPINE: a linear lifecycle (Intake → Plan → Build → Diffs → Monitor) where each stage is one full-bleed annotated real-product screenshot, one verb-noun label, and one sentence. The product's own pipeline IS the page's scroll order — the site literally walks you through the workflow in the order the work happens. Corroborated by dedicated stage pages (linear.app/intake, /plan).

★ Technique worth stealing (canon-safe)Make the page's scroll order BE the product's own lifecycle. For Kashi: a named-stage spine 'Record → Surface → Contest → Review' (one real /me screenshot + one verb label + one sentence per stage). It teaches a multi-step governance flow in one downward read without a single explanatory paragraph — and 'Contest/Review' bakes the human-review, contestable-signal canon INTO the spine.

⚠ do not copy: Do NOT copy the 'a new species of product tool' / 'self-driving' framing for Kashi's category claim — 'self-driving' adjacent to people-data reads as autonomous decisioning. Kashi must stay 'review-support signals for human review', never autonomous/self-driving over employee interaction data. The scrolly DEVICE is safe; the 'self-driving' WORD is not.

HeroHeadline (EN): "Financial Infrastructure to Grow Your Revenue" (JA live render: "事業成長を支える金融インフラ。") — Sub (JA live): "決済から組込み型金融、エージェンティックコマースまで、企業の収益最大化を支援。" (~ 'From payments to embedded finance to agentic commerce — helping businesses maximize revenue.')

How-it-works deviceLIVE CODE SAMPLE AS PROOF-OF-SIMPLICITY (canonical Stripe move): an editable/animated code block where a few lines stand in for 'integrate payments.' The 'how it works' is shown as 'how little it takes to work.' On the current homepage this is supported by a logo+case-study grid rather than a step diagram — the device is 'show the smallest working unit,' not 'diagram the steps.'

★ Technique worth stealing (canon-safe)The single load-bearing METAPHOR-NOUN that absorbs a sprawling product into one mental model. Kashi's safe analogue: position it as 'governance infrastructure for how meetings actually run' — a neutral, structural noun (like 'infrastructure'/'instrumentation') that frames repeated turn-taking/interruption asymmetries as plumbing-level signals for human review, not as judgments about people.

⚠ do not copy: Do NOT copy Stripe's two-co-equal-CTA pattern ('始める' + '営業にお問い合わせ' at equal weight) — Kashi needs ONE unmistakable next step. Also avoid Stripe's revenue/'maximize revenue/収益最大化' outcome-promise grammar applied to PEOPLE signals; for Kashi that grammar drifts toward scoring/evaluation. Borrow the infrastructure-noun framing, not the optimize-the-metric promise.

HeroHeadline: "Build and deploy on the AI Cloud." — Sub: "Vercel provides the developer tools and cloud infrastructure to build, scale, and secure a faster, more personalized web."

How-it-works deviceRESULT-METRIC-AS-PROOF + LIVE CODE SAMPLE: customer outcomes are placed as the immediate proof band right under the hero ('build times went from 7m to 40s' — Runway; '95% reduction in page load times' — Leonardo AI; '24x faster builds' — Zapier), and the AI Gateway concept is taught with a real TypeScript snippet (AI SDK streamText calling a model through one endpoint). The mechanism is shown by the smallest runnable code + the magnitude of the result.

★ Technique worth stealing (canon-safe)The QUANTIFIED-DELTA proof band directly under the hero: one-line, before→after, attributed customer results doing the persuading before any feature copy. Kashi-safe version: attributed pilot outcomes framed as process/review wins, e.g. 'surfaced 3 recurring interaction asymmetries the team chose to discuss in review' — concrete magnitude, attributed, with zero people-scoring or prediction language.

⚠ do not copy: Do NOT copy Vercel's stacked competing CTAs ('Deploy' + 'Get a Demo' + 'Talk to an Expert' + 'Get an Enterprise Trial' all on one page) — that CTA fork is the opposite of Kashi's needed single next step. (No people-data canon conflict in Vercel's framing itself; the issue is purely CTA proliferation + the generic 'AI Cloud' abstraction, which is vaguer than Kashi's hero should be.)

HeroHeadline: "High-performing teams are built here" / Sub: "Lattice is your team's daily destination for work, combining the best people and AI tools to help employees grow, managers lead, and companies succeed." CTAs: "Take a tour" + "Request a demo".

How-it-works deviceModular capability stack (named-pillar grid). The product is explained not by a 3-step flow but by a vertical stack of self-contained, benefit-titled modules — Performance, Goals & OKRs, Engagement, Grow, Compensation — each paired with an annotated product screenshot and a one-line outcome promise ("Survey smarter with AI-powered insights"). Below that, a second device: "Build high performance through everyday habits" reframes abstract HR software as a list of concrete recurring rituals (1:1s, Weekly Updates, Feedback, Q&A Boards), turning a platform into a daily behavior set.

★ Technique worth stealing (canon-safe)The "everyday habits / recurring rituals" anchor: explain Kashi by naming the concrete moments it attaches to (every recurring meeting, every 1:1, every standup) rather than by describing the analysis engine abstractly. Kashi-safe framing: "Kashi lives in the meetings your team already holds — it surfaces, after the fact, the repeated structural patterns in who speaks, who interrupts, and who returns to whom, as contestable signals for human review." This borrows the ritual-anchoring legibility move WITHOUT borrowing any scoring/grading language.

⚠ do not copy: Engagement is presented as a measurable, AI-analyzed module: corroborating product pages describe eNPS scoring, heatmaps, "sentiment tracking," "key driver analysis," and surfacing "engagement gaps." Kashi must NOT copy: (a) any "engagement"/eNPS scoring or sentiment-tracking framing, (b) the manager-leverage tilt ("help managers lead") since Kashi is whole-org not manager-only, (c) "27% attrition reduction" style outcome-prediction proof. Do NOT adopt sentiment/engagement scoring — canon conflict.

HeroHeadline: "The Performance Culture Quadrant™" / Sub: "Finally answer the question: Do we have a high performing culture?" CTAs: "Book a demo" + "Explore the PCQ". (Note: homepage also positions itself as "The market-leading employee experience platform.")

How-it-works device2x2 quadrant diagram (named, trademarked framework). The Performance Culture Quadrant™ plots two axes — Engagement (y) × Performance Confidence (x) — into four named culture archetypes (Peak Performance, Engaged Skepticism, Strained, Disconnected). The diagram itself IS the how-it-works: a visitor immediately understands "answer one question by finding which box you're in," then six "evidence-based dimensions" are presented as the levers that move you between boxes. Backed by a credibility device: "15+ years of research, 1.5 billion survey questions answered."

★ Technique worth stealing (canon-safe)Lead with the buyer's unspoken question, then hand them a single named artifact that answers it. For Kashi (canon-safe): hero could pose "Are the same people always getting cut off?" or "Whose questions go unanswered, meeting after meeting?" and answer with a named, content-blind STRUCTURAL artifact (e.g., a "return-to-whom map" / "turn-taking pattern review") — framed as a contestable signal for human review, never a score or a verdict. Steal the question-first + named-artifact legibility; reject the quadrant-scoring machinery entirely.

⚠ do not copy: MAJOR canon conflict — the entire hero is a scoring engine. PCQ™ is literally built on an "Engagement" index and "Performance Confidence" index; the platform self-describes as "an always-on intelligence layer (CultureOS™)" that turns culture into measurable indices, and case studies tout "11% increase in engagement scores." This is the textbook "engagement/health score" framing the lens warns about, plus prediction framing in their PR ("culture predicts market value"). Kashi must NOT adopt: engagement/health/culture scoring, a verdict 2x2 that labels a team "Disconnected," "always-on intelligence layer" (reads surveillance-adjacent), or any "predicts" claim. Do NOT adopt — canon conflict.

HeroHeadline: "The AI-powered performance management platform built for business impact" / Sub: "The AI-powered all-in-one people management system that's easy to use, delivers effortless insights, and enables managers to lead with impact". CTAs: "Book a Demo" + "See Pricing".

How-it-works deviceOutcome-funnel triad + dashboard-as-hero. The site organizes around the slogan "Insights. Action. Impact." — a 3-beat causal chain (you get insight → you act → you see business impact) that frames every module. The centerpiece device is the "HR Outcomes Dashboard," shown as a single command-center screen claiming to measure engagement, performance, turnover, and manager effectiveness in one view, so the dashboard screenshot itself does the explanatory work of "here's what you'd see."

★ Technique worth stealing (canon-safe)The compact causal-chain framing "signal → human action → outcome" as the site's organizing spine. Kashi-safe adaptation: "Pattern → conversation → change" — Kashi surfaces a repeated structural pattern, a human reviews and contests it, the team decides whether to change how meetings run. This borrows 15Five's legible cause-effect spine while keeping the human-in-the-loop and explicitly NOT promising a predicted business outcome or a score.

⚠ do not copy: MAJOR canon conflict — multiple forbidden mechanics at once. (1) MEI is an explicit continuously-updating "score" of an individual manager built from check-ins/1:1 activity — this is both scoring AND surveillance-adjacent behavioral tracking of a person. (2) "Predictive Impact Model" uses ML to "predict, with high confidence" engagement movement — prediction framing. (3) The whole site is manager-leverage-centric ("enables managers to lead"), conflicting with Kashi's whole-org stance. Kashi must NOT adopt: per-person effectiveness scores, "predict ... with high confidence," continuous behavioral signal harvesting from 1:1s, or manager-only positioning. Do NOT adopt — canon conflict.

HeroHeadline: "AI that works across your entire business" / Sub: "Connect global HR, IT, Finance, and Payroll in one platform, giving AI everything it needs to help run your company." CTAs: "Create free account" + "See a demo" + "Take a product tour".

How-it-works deviceSingle-source-of-truth architecture diagram ("unified data layer with people at the center"). Rather than steps, the explanatory device is a structural claim made visual: one data layer underneath, modular apps on top ("Start with the apps you need, or run them all together"), with the payoff line "When everything runs on one system, work moves faster." The architecture IS the pitch — the diagram shows data flowing through shared permissions and enterprise reporting, so the mechanism is shown as a system topology rather than a user journey.

★ Technique worth stealing (canon-safe)Ground the AI/analysis claim in an architectural primitive so it reads as earned, not magic. Kashi-safe analogue: lead with Kashi's defining primitive — "content-blind structural layer" — and let the value derive from it ("because Kashi reads only the structure of who-speaks-to-whom, never the words, it can surface repeated interaction asymmetries as auditable, reproducible signals"). This mirrors Rippling's "name the foundation, derive the rest" move and directly supports Kashi's privacy-bounded canon. (Use "auditable + reproducible + content-blind," never "deterministic.")

⚠ do not copy: Low direct conflict (Rippling is workforce ops/automation, not sentiment scoring) — but two things Kashi must NOT copy: (1) the "AI ... to help run your company" / "OS that's actually intelligent" all-encompassing-automation framing, which reads as decision-making-on-your-behalf and clashes with Kashi's human-review/contestable-signal stance; (2) the "unified data layer with people at the center" total-visibility framing — for a people-DATA product that phrasing tips toward surveillance, so Kashi should keep its "content-blind / privacy-bounded" qualifier loud whenever borrowing the architecture-primitive device. Adopt the device, not the omniscient-AI framing.

HeroHeadline: 「労務管理も、タレントマネジメントも」 (Both HR administration and talent management). Sub: 「業務効率化×戦略人事で事業を伸ばす」 (Grow the business through operational efficiency × strategic HR).

How-it-works deviceSystem-relationship diagram ('SmartHRのサービスと従業員データベースの関係図'): a single schematic showing that as you USE the labor-admin and talent modules, employee data accumulates automatically into one central DB ('入社から退社まで最新データを網羅的に収集'). The device teaches the product as a data-flywheel, not a feature list — the diagram itself carries the 'how'.

★ Technique worth stealing (canon-safe)The data-flywheel diagram: one schematic that shows the signal accumulating as a byproduct of normal use, so trust in the higher-order layer is *inferred from the mechanism*, not asserted. Kashi can adapt this canon-safely: a diagram showing that as meetings happen normally, structural interaction signals (turn-taking, who-returns-to-whom) accumulate content-blind into an auditable, reproducible record for human review — explaining the 'how' visually instead of claiming an outcome.

⚠ do not copy: Heavy 'No.1 / シェアNo.1' market-leadership badge stacking is identity-not-mechanism and not available to an early-stage product — do NOT copy the badge-wall approach as a primary hero proof. SmartHR's framing is otherwise canon-neutral (it never scores or predicts people).

HeroHeadline: 「個の力を最大化するタレントマネジメントシステム シェアNo.1 カオナビ」 (The talent-management system that maximizes individual potential — share No.1). Sub presents two resource CTAs rather than a sub-claim: 「3分でわかるカオナビ／いますぐ資料で確認してみたい」 and 「目的・企業規模別にわかる／詳しい料金を知りたい」.

How-it-works deviceBefore→After transformation strip backed by a face-photo-anchored product shot. The 'before' is '現場に眠っている個の人材情報' (individual talent data sleeping/scattered on the ground floor); the 'after' is an organized capability list (最適配置・育成計画・評価の最適化・スキルの可視化・予兆検知). The signature device is the 顔写真起点UI — a grid of employee face-photos as the literal entry point — which makes 'people data made visible' instantly legible without explaining a database.

★ Technique worth stealing (canon-safe)The 'invisible-today → made-visible-and-reviewable' before/after one-liner. Kashi's wedge is exactly an invisibility problem (repeated structural asymmetries no one can see across meetings). A canon-safe analog: 'これまで誰も見返せなかった会議の構造的な偏りを、人が確認できる形に' (what no one could revisit becomes something a human can review) — concrete before/after, no scoring, no prediction.

⚠ do not copy: Two phrases on this page are DIRECT canon violations Kashi must NOT copy: '予兆検知' (anomaly/sign DETECTION = prediction framing) and the '離職防止/attrition-prevention' use-case + AI '推奨アクション/課題提示' (an evaluation-and-recommend engine). These are prediction + scoring-adjacent. Flag: do NOT adopt '予兆検知', '離職予兆', or AI 'recommended action' framing for Kashi — replace with 'contestable review-support signals for human review'.

HeroHeadline: 「測定して『可視化』するだけでは、組織は変わらない。」 (Measuring and 'visualizing' alone won't change the organization.). Sub: 「Wevoxと『誰かがやってくれる』状態から脱却しましょう。」 (Break free, with Wevox, from the 'someone else will handle it' state.). A product-page variant also runs 「たった3分のパルスサーベイで組織の状態がわかる」.

How-it-works deviceNamed verb-cycle: 「たずねる・わかる・はなす・できる」 (Ask → Understand → Discuss → Do), mapped onto a STEP1 CHECK → STEP2 REVIEW → STEP3 ACTION loop, presented as a 'きづきサイクル' (awareness cycle). The whole product is taught as a repeating loop rather than a feature inventory — four plain-Japanese verbs make an abstract SaaS instantly graspable and imply continuity over time.

★ Technique worth stealing (canon-safe)The 'measurement/visualization is NOT the point — the human action after it is' hero stance, executed as a named plain-verb loop. This is the single most canon-perfect move for Kashi: it positions the tool explicitly as review-and-dialogue SUPPORT (not an answer engine), and a verb-cycle like 観る→確かめる→話す→見直す (observe → contest → discuss → revisit) frames the signals as contestable inputs for human review — zero scoring, zero prediction, whole-org dialogue framing.

⚠ do not copy: Wevox's OWN top-of-funnel is canon-safe (it avoids スコア and leads with dialogue), but its ECOSYSTEM/partner and affiliate copy frames the product as 'エンゲージメントを把握して離職を防げ' (grasp engagement to PREVENT attrition) and leans on 'エンゲージメントスコア'. Do NOT adopt: 'engagement score', 'パルスサーベイで離職を防ぐ', or any attrition-prevention promise. Borrow the dialogue-loop stance, never the score/attrition layer.