saglitzdesign-mcp 0.3.1

package/knowledge/marketing/ad-creative.md

@@ -0,0 +1,155 @@
+---
+id: ad-creative
+title: "Ad Creative — Hooks, Formats, Specs & Testing"
+category: marketing
+platform: both
+tags: [ads, creative, meta, tiktok, google, ugc, testing]
+sources: ["https://www.tryvizup.com/blog/meta-ad-specs-2026-every-dimension-size-you-need", "https://billo.app/blog/meta-ads-safe-zones/", "https://support.google.com/google-ads/answer/7684791", "https://quickframe.com/blog/youtube-video-ad-specs", "https://ads.tiktok.com/help/", "Breakthrough Advertising (Eugene Schwartz)", "Ogilvy on Advertising (David Ogilvy)"]
+updated: 2026-07-08
+---
+
+# Ad Creative — Hooks, Formats, Specs & Testing
+
+## Creative Is the Targeting Now
+
+- Post-ATT, with broad targeting and AI-driven delivery (Advantage+ / Smart+ / PMax), **the algorithm finds the audience through the creative.** Who the ad visually and verbally speaks to determines who it's shown to. Creative quality/diversity is the #1 performance lever an advertiser controls.
+- Practical consequence: make the intended customer unmistakable in the first seconds ("If your team still plans sprints in spreadsheets…") — you're targeting via the hook.
+- Budget implication: expect to produce and test many concepts continuously; media buying is increasingly automated, creative production is the moat.
+
+## Hook-First Structure (first 1–3 seconds decide everything)
+
+- **~50%+ of skips happen in the first 3 seconds. The hook is the ad.** Structure video as: HOOK (0–3s) → PROBLEM/AGITATE (3–10s) → SOLUTION/DEMO (10–20s) → PROOF (20–25s) → CTA (last 3–5s).
+- Hook rules:
+  - Open mid-action or mid-sentence — never with a logo, title card, or slow establishing shot.
+  - Visual hook + verbal hook + text overlay hook simultaneously (many watch muted; always burn in captions).
+  - Show the product or the problem within the first 2 seconds.
+  - Pattern interrupt: unexpected visual, bold claim, direct question, "stop scrolling if…".
+- Hook formulas: negative call-out ("Stop doing X"), result-first ("This took our CAC from $80 to $31"), POV/UGC ("I tested every X so you don't have to"), question ("Why does nobody talk about…?"), before-state shock (the messy desk, the error screen).
+- Test 3–5 hooks per body: swapping only the first 3 seconds is the cheapest high-impact creative test.
+
+## Platform Formats & Current Specs (verified mid-2026 — re-check before big buys)
+
+### Meta (Facebook/Instagram)
+- **Feed:** 1:1 (1080×1080) or 4:5 (1080×1350 — Meta's recommended feed default; ~1% CTR lift vs 1:1 for images). Primary text ~125 chars visible, headline ~40 chars before truncation.
+- **Stories/Reels:** 9:16, 1080×1920 (export 1440×2560 for quality). **Unified 2026 safe zones: keep critical content out of top ~14%, bottom ~20–35%, and ~6% each side** (UI overlays: username top, CTA/engagement bottom). 9:16 video shows ~7% higher CTR vs. letterboxed.
+- Video: MP4/MOV, ≤ 4 GB; keep 15–30s for feed, ≤ 15s preferred for stories/reels. Provide 1:1 + 4:5 + 9:16 renditions of every concept (placement asset customization).
+- Sound-on design but caption everything.
+
+### TikTok
+- 9:16, 1080×1920, MP4/MOV, 5–60s allowed; **21–34s is TikTok's cited sweet spot; hooks judged in first 1–2s.** Safe zones: avoid bottom ~130 px+ (caption/CTA) and right rail ~120 px (engagement icons).
+- Native-or-die: shot-on-phone look, creator voice, trending audio/format references, captions burned in. Polished TVC-style creative reliably underperforms here. Spark Ads (boosting creator posts) usually beat dark-side uploads of the same asset.
+
+### YouTube (Google)
+- **Bumper:** 6s non-skippable — one message, no arc, logo/brand visible throughout.
+- **Skippable in-stream:** skip button at 5s → your brand and value proposition must land inside the first 5 seconds. Recommended 15–30s (12s min, up to 3 min). 1920×1080 MP4.
+- **Shorts ads:** 9:16 vertical, treat like TikTok.
+- In-feed/display companions: 16:9 + 1:1 thumbnails.
+
+### Google Search & Display
+- **Responsive Search Ads:** up to 15 headlines × 30 chars, 4 descriptions × 90 chars, 2 path fields × 15 chars. Provide the max assets; include keywords in ≥ 5 headlines; pin sparingly (pinning cuts combinations Google can optimize). Google serves up to 3 headlines + 2 descriptions per auction.
+- **Responsive Display:** landscape 1200×628 (1.91:1), square 1200×1200, logo 1200×1200 + 1200×300; headlines 30 chars, long headline 90, descriptions 90. Text must occupy < 20% of image area for best serving.
+- **Demand Gen:** reuse 1:1 / 4:5 / 9:16 social assets.
+
+### App Store / Play ads
+- **Apple Search Ads:** creative is pulled from App Store product page assets — screenshots/previews ARE the ad; Custom Product Pages let you match creative to keyword themes (e.g., "budget app" keyword → budgeting screenshot set first).
+- **Google App Campaigns:** supply text (30/90 char), images (1200×628, 1200×1200), video (16:9, 1:1, 9:16), and playables; the system mixes assets — every asset must stand alone.
+
+## UGC-Style vs. Polished Brand Creative
+
+- **UGC-style** (creator-shot, handheld, talking-to-camera, "honest review" energy): best for cold prospecting on TikTok/Reels, low-consideration products, demonstrating use in real life, and testing many angles cheaply. Wins on native feel and trust, not production value. Must still follow disclosure rules (#ad) and can't fake being an organic review.
+- **Polished brand creative** (studio, motion design, cinematic): best for retargeting/consideration stages, premium positioning, brand campaigns, YouTube in-stream, and categories where polish signals trust (finance, health, luxury, B2B).
+- Default playbook: **prospect with UGC-style volume, retarget and brand-build with polish.** Blend formats in one account — creative diversity itself improves delivery.
+- "UGC-style" ≠ low effort: it's scripted native creative — hook, demo, proof, CTA — wearing casual clothes.
+
+## Static Ad Layout Formulas
+
+1. **Problem/Solution split:** left/top = pain (visual of the before-state, red/muted), right/bottom = product resolving it (clean, brand color). Headline names the pain, subhead names the fix.
+2. **Us-vs-Them comparison table:** your brand column (✓✓✓, brand color) vs. "other ways" column (✗, gray). 3–5 rows, benefit-phrased. High performer for challenger brands; keep claims provable.
+3. **Testimonial card:** big pull-quote (result-specific: "Cut our reporting time 6 hours/week"), customer photo + name + title, small product shot, star rating if real. Quote IS the headline.
+4. **Feature callout / annotated product:** product hero shot with 3–4 thin callout lines labeling concrete benefits. Works for gadgets, apps (annotated screenshot), CPG.
+5. **Bold claim + proof:** giant short headline (≤ 7 words), one supporting stat line, product, CTA. The "billboard" — best when the offer itself is strong.
+6. **Meme/native text-post format:** looks like organic platform content; strong for cold social, off-brand for premium.
+- Universal statics rules: one message per ad; headline legible at thumbnail size; product visible; logo small and NOT the hero; CTA present but not dominant; 4:5 master crop.
+
+## Ad Copy Formulas
+
+- **Hook–Benefit–Proof–CTA (default):**
+  1. Hook: call out audience or pain ("Running Meta ads with a 3-person team?")
+  2. Benefit: outcome, not feature ("Launch tested creative twice as fast")
+  3. Proof: number, testimonial, or credential ("2,400 brands", "4.8★")
+  4. CTA: one action + friction remover ("Start free — no card required")
+- Others in rotation: PAS (Problem–Agitate–Solution) for pain-aware audiences; AIDA for unaware; "How [persona] gets [result] without [pain]"; social-proof-first ("2,400 teams switched…").
+- Awareness matching (Schwartz): unaware → lead with the problem story; problem-aware → lead with the mechanism; solution-aware → lead with differentiation; product-aware → lead with offer/deal.
+- Write copy at a 6th–8th-grade reading level; front-load meaning (feeds truncate); numbers beat adjectives.
+
+## Message Match (Ad → Landing Page "Scent")
+
+- **The landing page must repeat the ad's promise in its headline — ideally near-verbatim.** Same offer, same numbers, same visual style, same hero person/product where possible.
+- Mismatch (ad promises "50% off first month", page shows generic homepage) is the #1 silent conversion killer and raises paid-search costs via lower Quality Score/relevance signals.
+- Rules: one landing page per angle/offer (not per campaign); ad's CTA verb = page's button verb; keep the discount/claim above the fold; UGC ads → pages with matching UGC/testimonials perform better than corporate pages.
+- Audit method: screenshot ad + first viewport of landing page side by side — a stranger should instantly see they're the same story.
+
+## Creative Testing Method
+
+- **Volume first:** launch distinct *concepts* (different angle/hook/format), not micro-variants. Baseline: 3–5 new concepts per week for scaled accounts; minimum 2–4/month for small ones.
+- Test structure: broad audience, one variable per test at the concept level; let the platform's dynamic systems handle micro-optimization. Judge on CPA/ROAS and hook rate + hold rate (3-sec views ÷ impressions; ThruPlay ÷ 3-sec) — not CTR alone.
+- Statistical hygiene: decide on a spend threshold per variant (e.g., 1–2× target CPA spent) before killing; don't call winners in the first 24h (learning phase noise).
+- **Iterate on winners, don't reinvent:** when a concept wins, spawn variants — new hooks on same body, new format of same angle (video→static→carousel), new creator/voice, new first frame. A winning *angle* is the asset; exhaust it.
+- **Fatigue signals:** frequency creeping > 2–4 (prospecting), CTR down ~20–30% from peak, rising CPA with flat everything else, hook rate decay. Response: rotate in fresh concepts/hooks — refreshing the first 3 seconds often resets fatigue at a fraction of the cost of new production.
+- Keep an always-on library: tag every asset by angle, hook type, format, creator; log results so learnings compound.
+
+## Compliance Basics
+
+- **Claims must be substantiated:** any specific performance, income, health, or savings claim needs proof you hold *before* running (studies, verified data). "Results not typical" doesn't fix an unrepresentative claim (FTC).
+- **Before/after imagery:** restricted or banned for health/weight-loss on Meta and TikTok; must not promote negative self-perception. Same-lighting/pose honesty required where allowed.
+- **Endorsements:** creators must disclose material connection (#ad — clearly, not buried); testimonials must reflect real, typical experience; fake reviews are illegal (FTC rule, 2024+).
+- Restricted categories with special rules/verification: housing, employment, credit (Meta: no age/gender/zip targeting), health/pharma, finance/crypto, alcohol, gambling, political.
+- Platform-specific: no excessive skin/shock imagery, no "you"-statements implying knowledge of personal attributes ("Are you diabetic?"), no fake UI elements (fake play buttons/notifications), music licensing — commercial-use libraries only for ads (trending organic sounds are usually NOT cleared for paid).
+
+## Creative Diagnostics (which metric blames which part of the ad)
+
+- **Low hook rate** (3-sec views ÷ impressions, target > 25–30%): the first 3 seconds fail — new hook, not new ad.
+- **Good hook rate, low hold rate** (ThruPlay/15s ÷ 3-sec views): body drags — tighten pacing, cut to demo sooner.
+- **High CTR, low conversion rate:** ad over-promises or landing page mismatch — fix scent before killing the creative.
+- **Low CTR, fine conversion rate:** weak CTA/offer framing on a qualified audience — test CTA and end-card, keep the angle.
+- **Everything fine, CPA rising over weeks:** fatigue — rotate hooks/refresh concepts, check frequency.
+
+## Quick-Reference Spec Cheat Sheet (mid-2026)
+
+| Placement | Ratio | Export size | Length / limits | Critical rule |
+|---|---|---|---|---|
+| Meta Feed (image) | 4:5 | 1080×1350 | primary text ~125 chars visible | 4:5 beats 1:1 for images |
+| Meta Feed (video) | 4:5 or 1:1 | 1080×1350 | 15–30s | captions burned in |
+| Stories/Reels | 9:16 | 1080×1920 (1440×2560 best) | ≤ 15s preferred | safe zone: top 14%, bottom 20–35%, sides 6% |
+| TikTok | 9:16 | 1080×1920 | 21–34s sweet spot | native look; hook in 1–2s |
+| YouTube skippable | 16:9 | 1920×1080 | 15–30s (skip at 5s) | brand + value < 5s |
+| YouTube bumper | 16:9 | 1920×1080 | 6s max | one message only |
+| YouTube Shorts | 9:16 | 1080×1920 | < 60s | treat as TikTok |
+| Google RSA | text | — | 15×30-char headlines, 4×90-char descriptions | max assets, pin sparingly |
+| Google Display | 1.91:1 + 1:1 | 1200×628, 1200×1200 | headlines 30/90, desc 90 | text < 20% of image |
+| App campaigns | 16:9, 1:1, 9:16 | 1920×1080 etc. | every asset standalone | store assets ARE the ad (ASA) |
+
+Specs shift quarterly — confirm against platform docs before a major flight.
+
+## Pre-Launch Creative Checklist
+
+1. Hook lands in first 2 seconds (visual + verbal + text overlay); no logo intro.
+2. Comprehensible on mute; captions burned in; safe zones respected per placement.
+3. Native renditions exported: 9:16, 4:5, 1:1 (no letterboxing).
+4. One message, one CTA; audience call-out explicit (creative-as-targeting).
+5. Landing page headline mirrors ad promise; offer/numbers identical; CTA verbs match.
+6. Claims substantiated and documented; disclosures (#ad) present; music licensed for paid.
+7. Naming convention applied (angle_hook_format_date) so results feed the learning library.
+8. Kill/scale thresholds decided in advance (spend = 1–2× target CPA per variant).
+
+## Anti-Patterns
+
+- **Logo-first videos / 3-second brand stings before content** — you pay to be skipped.
+- **30-second builds:** saving the point for the end of a skippable ad. Front-load; assume every second is the last one watched.
+- **Ad-promise ↔ landing-page mismatch** (different offer, headline, or visual world).
+- One 16:9 asset letterboxed everywhere instead of native 9:16/4:5/1:1 renditions; ignoring safe zones so CTAs sit under platform UI.
+- Sound-dependent creative without captions; hooks that require audio to make sense.
+- Testing 10 shades of button color while running one creative concept; killing ads in 24h; never killing anything.
+- Polished TVC recuts pushed to TikTok; fake-UGC that pretends to be organic (trust + policy risk).
+- "We tried video once" — treating creative as a launch task instead of a weekly production system.
+- Claim inflation ("best", "#1", "guaranteed") without substantiation — account bans outlast any CTR bump.

package/knowledge/marketing/branding-identity.md

@@ -0,0 +1,153 @@
+---
+id: branding-identity
+title: "Branding & Identity — Strategy to Logo System"
+category: marketing
+platform: both
+tags: [branding, logo, identity, brand-voice, naming]
+sources: ["The Brand Gap (Marty Neumeier)", "Zag (Marty Neumeier)", "A Designer's Art (Paul Rand)", "How Brands Grow (Byron Sharp)", "Designing Brand Identity (Alina Wheeler)", "https://developer.apple.com/design/human-interface-guidelines/app-icons", "https://developer.android.com/distribute/google-play/resources/icon-design-specifications"]
+updated: 2026-07-08
+---
+
+# Branding & Identity — Strategy to Logo System
+
+## Strategy Before Style (Neumeier)
+
+- **A brand is not a logo.** A brand is a person's gut feeling about a product, company, or service. You don't own the brand — customers do. Design only shapes the inputs to that feeling.
+- **Do strategy first, always.** Never start sketching logos until you can answer: Who is this for? What do they get? Why should they believe it? Who else offers it?
+- **Write the onlyness statement (Zag) before any visual work:**
+  > "Our [offering] is THE ONLY [category] that [key differentiator] for [audience] in [market/geography] in an era of [underlying trend]."
+  If you can't fill in "THE ONLY," the differentiator is too weak — narrow the category or sharpen the difference until it's true.
+- **Radical differentiation beats incremental improvement.** When everybody zigs, zag. If competitors are all blue and corporate, that is permission — not a mandate — to be warm and human. Differentiate on strategy first; visuals express the zag, they don't create it.
+- **Three questions test every brand decision:** Is it different? Is it relevant? Is it credible? All three must be yes.
+
+## Logo Design Principles (Paul Rand)
+
+- **Simplicity wins.** A logo's job is identification, not explanation. It does not need to show what the company does (Apple sells no fruit; Shell sells no shells).
+- **Rand's test order:** Is it distinctive? Is it memorable? Is it appropriate? Appropriateness means fitting the brand's character — not literal depiction of the product.
+- **Avoid literalism.** A camera icon for a photography studio is a category label, not a brand. Prefer abstract or lateral marks that acquire meaning through use.
+- **The 16px-to-billboard test (mandatory):** Render the mark at 16×16 px (favicon) and imagine it 10 m wide. It must be recognizable at both. If detail disappears at 16px, remove the detail from the design — do not create a "simplified version" as an afterthought; simplify the master.
+- **Design in black first.** If a mark doesn't work in one flat color, color is doing the work the form should do. Color is applied after the form is proven.
+- **A logo derives meaning from the quality of the thing it symbolizes, not the other way around.** Don't over-invest in symbolism narratives; invest in distinctiveness and craft.
+
+## Logo System Deliverables (Minimum Viable Identity)
+
+Every identity handoff must include:
+
+1. **Primary lockup** — wordmark + mark in fixed relationship.
+2. **Standalone wordmark** — for contexts where the mark adds noise.
+3. **Standalone mark/symbol** — for avatars, favicons, app icons.
+4. **Horizontal and stacked lockups** — wide headers vs. square placements.
+5. **Clear space rule** — defined by an element of the logo itself (e.g., "clear space = height of the letter X on all sides"), never in absolute pixels.
+6. **Minimum sizes** — specify separately for print (mm) and screen (px). Typical: wordmark min 80–100 px wide on screen; standalone mark min 16 px.
+7. **One-color (mono) version** — solid black and solid white, no gradients or transparency. Required for engraving, embroidery, stamps, single-color print.
+8. **Reversed (knockout) version** — tested on the brand's darkest color, on photography, and on black.
+9. **File package** — SVG (master), PDF (print), PNG at 1x/2x/4x with transparency, and .ico/favicon set.
+
+**Rule:** if any of these versions requires redrawing shapes (not just recoloring), the master logo is over-detailed — fix the master.
+
+## Logo → App Icon Translation
+
+An app icon is not a logo dropped in a square. Rules:
+
+- **Use the standalone mark, never the full wordmark.** Text is illegible at icon sizes; a name under an icon is already provided by the OS.
+- **Fill the canvas.** Logos are designed with breathing room; icons should use ~70–80% of the tile. Scale the mark up beyond what feels comfortable in the guidelines doc.
+- **iOS:** design at 1024×1024 px, square; the system applies the superellipse mask — never pre-round corners or add your own drop shadow. Since iOS 18+, provide light, dark, and tinted variants (dark = mark on dark/transparent background; tinted = grayscale mark the system tints). Liquid Glass era (iOS 26+): prefer layered, flat vector shapes via Icon Composer so the system can apply material effects.
+- **Android:** adaptive icons = 108×108 dp canvas with foreground and background layers; only the inner 66 dp is guaranteed visible (masks vary: circle, squircle, rounded square). Keep the mark inside the 66 dp safe zone. Play Store listing icon: 512×512 px, no transparency.
+- **Monochrome/themed variant (Android 13+):** supply a single-color foreground layer for themed icons.
+- **Test the icon on a real home screen next to the top-10 apps.** If it disappears among them, increase contrast or simplify.
+
+## Color & Typography as Brand Assets (Byron Sharp)
+
+- **Distinctive assets > differentiated messaging.** Sharp's evidence: brands grow through mental and physical availability. Visual assets (color, shape, type, character, sound) work by making the brand instantly recognizable, not by communicating positioning.
+- **Pick ONE ownable color and use it with extreme consistency.** Category examples: Tiffany blue, Cadbury purple, McDonald's red-yellow. A palette of six equal colors builds nothing; a hierarchy of one dominant + supporting neutrals builds an asset.
+- **Asset criteria: fame and uniqueness.** Before committing, ask: (1) Will heavy repetition make this famous among buyers? (2) Is it unique in the category (not the world)? A blue fintech logo fails uniqueness; audit the top 10 competitors' colors before choosing.
+- **Typography:** choose one distinctive display face (headlines, logo-adjacent) + one workhorse text face. The display face carries brand recognition; the text face carries usability. License for web + app + print from day one.
+- **Never redesign distinctive assets for freshness.** Every change spends accumulated memory. Evolve assets only when they actively fail (illegibility, technical constraints), and keep the recognizable skeleton.
+
+## Brand Voice Definition
+
+Deliver voice as three artifacts:
+
+1. **3–4 voice attributes**, each with a boundary: "Confident, but never arrogant." "Playful, but never silly." "Plain-spoken, but never blunt." The "but never" clause is what makes an attribute usable.
+2. **Tone spectrum** — voice is constant, tone flexes by context. Map contexts on a spectrum: marketing page (most expressive) → onboarding → transactional email → error message → legal/security incident (most neutral). Rule: the higher the user's stress, the flatter the tone.
+3. **Do/don't pairs** (minimum 5) — real sentences, not adjectives:
+   - Don't: "Oops! Something went wrong 🙈" → Do: "We couldn't save your changes. Retry, or copy your text first."
+   - Don't: "Leverage our best-in-class solution" → Do: "Ship your first campaign in 10 minutes."
+
+## Naming Basics
+
+- **Name types, in rough order of ownability:** invented (Kodak, Zerply) > altered real word (Netflix, Lyft) > metaphor (Amazon, Slack) > descriptive (General Motors) > acronym (worst: zero meaning, zero memorability).
+- **Screening checklist (all required):** ≤3 syllables preferred; easy to spell after hearing it once ("radio test"); no negative meaning in major target-market languages; trademark class search done; domain or acceptable modifier available (get-, use-, .app, .io); social handles available or close; doesn't box in future scope ("BostonPizza problem").
+- **Descriptive names trade memorability for explanation — acceptable only for features and sub-products, not the master brand.**
+- Sub-brand rule: default to branded house ("Brand + descriptor": Google Maps, Stripe Atlas) unless the sub-offering targets a conflicting audience or risk profile — only then use a separate name.
+
+## Brand Guidelines Document Structure
+
+Order the document by decision frequency, not ceremony:
+
+1. Brand strategy on one page (onlyness statement, audience, attributes)
+2. Logo system (versions, clear space, min sizes, misuse examples)
+3. Color (values in HEX/RGB/CMYK/Pantone, usage ratios, accessibility pairings with contrast ratios)
+4. Typography (faces, weights, scale, fallback stacks)
+5. Voice & tone (attributes, spectrum, do/don't)
+6. Imagery & illustration style (with good/bad examples)
+7. Applications (social templates, deck, email signature, merch)
+8. Asset links (single source-of-truth folder, versioned)
+
+- **Misuse examples are the most-used page.** Show ≥6 forbidden treatments: stretched, recolored, outlined, drop-shadowed, on busy photo, rotated.
+- Keep it live (web page or Figma), not a PDF that forks into stale copies.
+
+## Brand Architecture (choosing the model)
+
+- **Branded house** (one master brand: Google, FedEx): default choice. Cheapest to build, every product deposit compounds into one brand. Use unless there's a hard conflict.
+- **House of brands** (P&G model, separate brands): only when audiences, price tiers, or risk profiles genuinely conflict (a luxury line and a discount line; a consumer app and an enterprise platform sold against each other).
+- **Endorsed** ("X by Brand"): transitional or trust-lending model — good for acquisitions keeping equity, or new categories that need the parent's credibility.
+- Decision rule: every new brand you create multiplies marketing cost and divides memory. Require written proof of audience conflict before approving a second brand.
+
+## Color Pairing & Accessibility Rules
+
+- Ship the palette WITH pre-approved text/background pairings and their contrast ratios; don't make every designer re-derive them.
+- Body text pairings must hit **4.5:1** contrast (WCAG AA); large text and UI icons **3:1**. Test the brand color as a button background with white AND dark text — pick one and standardize.
+- Define usage ratio explicitly (e.g., **60% neutral / 30% secondary / 10% brand accent**). The brand color stays scarce so it stays distinctive.
+- Provide dark-mode equivalents for every brand color at spec time (usually desaturated + lightened accents, near-black surfaces) — retrofitting dark mode breaks palettes.
+- Never encode meaning in brand color alone (success/error need their own semantic colors, distinct from the brand accent).
+
+## Rebrand vs. Refresh Decision Rules
+
+**Refresh** (evolve logo/type/color, keep name and recognizable assets) when:
+- Assets fail technically (won't scale, illegible on screens, dated rendering effects like bevels/gloss)
+- Visual system is inconsistent but the brand has positive recognition
+- Audience or product expanded but perception is neutral-to-positive
+
+**Rebrand** (new name and/or fundamentally new identity) only when:
+- Legal/trademark conflict forces it
+- Merger/acquisition changes the entity
+- The name blocks growth (wrong geography, category, or connotation)
+- The brand carries strong negative equity that no messaging can fix
+
+**Never rebrand because:** leadership is bored, a new CMO arrived, competitors refreshed, or "it feels dated" without user evidence. Test first: if unaided recognition of current assets is high, a rebrand destroys paid-for memory (Sharp). Budget rule: a rebrand costs 5–10× the design fee in rollout (signage, packaging, ads re-teaching the market) — price that before deciding.
+
+## Logo Evaluation Checklist (score a candidate before presenting)
+
+Run every candidate mark through all ten; any "no" sends it back:
+
+1. Recognizable at 16×16 px favicon?
+2. Works in solid black, one color, no gradients?
+3. Works reversed (white) on the darkest brand color and on photography?
+4. Distinct from the top 10 competitors' marks at a glance (do the lineup test)?
+5. Describable in one sentence over the phone? ("A lowercase g whose tail forms an arrow")
+6. No trend-locked styling (current gradient fads, glass effects, AI-swirl geometry)?
+7. Legible/appropriate across all target cultures and scripts?
+8. Free of accidental shapes (rotate it, mirror it, squint at it — offensive or unfortunate silhouettes)?
+9. Survives cheap reproduction (embroidery, stamp, laser engraving, photocopier)?
+10. Trademark screening in relevant classes passed?
+
+## Anti-Patterns
+
+- **Trend-chasing logos:** blanding into the same geometric sans wordmark as everyone else; adopting the gradient-of-the-year, glassmorphism, or AI-swirl marks. Trends guarantee your identity expires with the trend and looks like competitors' — the opposite of a brand's job.
+- **Unscalable detail:** thin lines, tiny text, photographic textures, 4+ color gradients inside the mark. If it needs a "small-size version" with different shapes, the master failed the 16px test.
+- **Logo-by-committee:** averaging feedback removes every distinctive edge. Present 2–3 strategically distinct directions, decided against the onlyness statement, not personal taste.
+- **Meaning-stuffing:** cramming five symbolic references into one mark ("the negative space is a bird AND our founder's initials AND..."). Memorability comes from simple form, not layered symbolism.
+- **Palette sprawl:** 8 "brand colors" of equal weight = no color asset. Enforce a dominance hierarchy (e.g., 60/30/10).
+- **Guidelines without misuse examples** and **identities shipped without mono/reversed versions** — both guarantee brand erosion within a year.
+- **Rebranding the assets that were working:** changing the one thing customers recognized (color, mascot, wordmark silhouette) while keeping everything customers ignored.

package/knowledge/marketing/email-marketing.md

@@ -0,0 +1,155 @@
+---
+id: email-marketing
+title: "Email Marketing — Design, Lifecycle Flows & Deliverability"
+category: marketing
+platform: web
+tags: [email, newsletter, lifecycle, deliverability, design]
+sources: ["https://www.litmus.com/blog/the-ultimate-guide-to-dark-mode-for-email-marketers", "https://www.emailonacid.com/blog/article/email-development/dark-mode-for-email/", "https://www.klaviyo.com/uk/blog/email-marketing-benchmarks-open-click-and-conversion-rates", "https://www.brevo.com/blog/email-marketing-benchmarks/", "https://mailchimp.com/resources/email-marketing-benchmarks/", "Email Marketing Rules (Chad S. White)"]
+updated: 2026-07-08
+---
+
+# Email Marketing — Design, Lifecycle Flows & Deliverability
+
+## Design Constraints (2026 Reality)
+
+- **Width: 600–640 px max.** Single column by default. Gmail clips messages over ~102 KB of HTML ("[Message clipped] View entire message") — keep HTML weight under 100 KB.
+- **Layout: tables are still the safe skeleton.** Outlook desktop (2016/2019/2021 and classic desktop) renders with Microsoft Word's engine: no flexbox, no grid, no `max-width` on divs, unreliable padding on `<div>`. It's a dual-Outlook world — "new" Outlook is Chromium-based, classic is Word-based — so code hybrid/"spongy" layouts: table-based structure with `<!--[if mso]>` conditional comments for Word-engine fixes, progressive enhancement (media queries, flexbox) for everything else.
+- **Outlook quirk checklist:** use `mso-line-height-rule: exactly`; set explicit widths on table cells; background images need VML fallback; border-radius is ignored (design buttons that survive square corners); animated GIFs show only the first frame in classic Outlook — put the message in frame 1.
+- **Inline your CSS** (or use a pre-send inliner). `<style>` blocks are supported by major clients now but still stripped in some contexts (forwarding, some international webmail); inline is the fallback that always works.
+
+### Dark Mode (design for it, don't fight it)
+
+Clients handle dark mode three ways: no change (leave as-is), partial inversion (invert light backgrounds only), full inversion (Gmail app on some platforms, Outlook apps — flips your colors even if you designed dark). Rules:
+
+1. Add the meta + CSS opt-in so clients that respect it use your styles:
+   `<meta name="color-scheme" content="light dark">` and `:root { color-scheme: light dark; }`, then override with `@media (prefers-color-scheme: dark)` (fully honored by Apple Mail; partially elsewhere).
+2. **Logos and key images: transparent PNG with a thin white/light stroke or outer glow** so a dark logo survives on dark backgrounds. Never bake a white rectangle behind the logo.
+3. **Never use pure black (#000) or pure white (#fff)** — near-neutrals (#111–#1a1a1a, #f5f5f7) invert more gracefully and reduce ugly forced-inversion artifacts.
+4. Avoid large solid background images with embedded text — inversion can make text unreadable and you can't override it.
+5. Test in: Apple Mail (respects your CSS), Gmail apps (partial/full invert, ignores your dark CSS), Outlook mobile (full invert). Design so the *inverted* version is acceptable, not just the designed one.
+
+### Images & Fonts
+
+- **Live-text rule: every critical message must survive with images off.** Some clients and privacy proxies block or delay images; all-image emails show a blank rectangle and scream "spam" to filters. Headline, offer, and CTA must be HTML text.
+- Meaningful `alt` text on every image; styled alt text (`font-family`, `color` on the `<img>`) as graceful degradation.
+- Retina images: export at 2x, constrain with width attribute.
+- **Fonts: web fonts render reliably only in Apple Mail and some webmail.** Always declare a full system stack fallback, e.g. `font-family: 'BrandFont', -apple-system, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif;`. Design must look intentional in the fallback, because most recipients see the fallback.
+- Minimum text sizes: 16 px body, 22 px+ headlines, 14 px absolute floor for legal text. Tap targets ≥ 44×44 px.
+
+## Subject Line + Preheader
+
+- **Subject: 30–50 characters** (mobile truncates ~30–40). Front-load the key word; the end gets cut, not the start.
+- **Preheader: 40–100 characters that CONTINUE the subject, never repeat it.** Never let "View this email in your browser" be the preheader. Kill leftover space with `&zwnj;&nbsp;` padding only after a real preheader.
+- One idea per subject. Specificity beats cleverness; curiosity beats completeness (open = wanting to know more).
+- **Formulas that work:**
+  - Benefit, plain: "Ship your first campaign in 10 minutes"
+  - Number + payoff: "3 layouts that doubled our reply rate"
+  - Curiosity gap (must be paid off inside): "The one metric we stopped tracking"
+  - Urgency, honest only: "Ends tonight: 20% off annual plans"
+  - Question the reader already asks: "Is your welcome email losing money?"
+  - Personal/social proof: "How Arda cut churn 18%"
+- **Bans:** ALL CAPS, more than one emoji, "RE:"/"FWD:" fakery, clickbait the body doesn't pay off (kills long-term opens), spam-trigger stuffing ("FREE!!!", "$$$", "Act now").
+- A/B test subjects on ≥1,000 recipients per arm; judge winbacks/promos on clicks or conversions, not opens (Apple MPP inflates opens).
+
+## Lifecycle Flows (sequence, timing, content)
+
+**Automated flows out-earn campaigns (roughly 3× click rate, ~13× order rate per Klaviyo data). Build these four first.**
+
+### Welcome (trigger: signup) — highest open rate you'll ever get (50%+)
+1. **Immediately:** deliver the promised thing (discount/lead magnet), one-line brand promise, single CTA. Set expectations (what you'll send, how often).
+2. **Day 2–3:** brand story or best-of content; social proof.
+3. **Day 5–7:** product/category education or objection-handling; soft CTA.
+Rule: no hard-sell before email 1 delivers value. 3–5 emails total.
+
+### Onboarding (trigger: account created / trial start) — goal: activation, not engagement
+- Map emails to activation milestones, not the calendar: "created project → invite team → first publish."
+- Each email = one action, one CTA, one screenshot/GIF of the action.
+- **Behavioral branching:** skip or swap emails the user has already completed. Never ask someone to do what they've done.
+- Cadence: day 0 (welcome/first action), day 1–2 (core action), day 3–5 (aha feature), day 7 (social proof + upgrade path), trial-end minus 3 days (what you lose), trial end (clear decision email).
+
+### Cart Abandonment (trigger: cart with email, no purchase)
+1. **+1–4 hours:** pure reminder — cart contents with images, one "Return to cart" CTA. No discount.
+2. **+24 hours:** objection handling — shipping/returns/guarantee, reviews of the carted item.
+3. **+48–72 hours:** incentive if margins allow (discount or free shipping), with a real expiry. Offering the discount in email 1 trains discount-waiting.
+Browse-abandonment variant: same skeleton, softer copy ("Still looking at X?"), 1–2 emails only.
+
+### Winback (trigger: no opens/clicks/purchases in 60–90 days B2C, 90–180 days SaaS)
+1. "We miss you" + best-value summary of what changed.
+2. Incentive or concrete new-feature hook.
+3. **Breakup email:** "We'll stop emailing you" with one-click stay button. Then actually suppress non-responders — this protects deliverability and is not optional.
+
+## Newsletter Design Patterns
+
+- Pick one architecture and keep it stable so scanning becomes habitual:
+  - **Single-story:** one essay/feature, one CTA — highest engagement per topic.
+  - **Digest:** 3–5 items, each = bold headline + 1–2 sentence summary + link. Ruthless hierarchy: lead item gets image + space; the rest are compact.
+  - **Curation:** links with one-line "why this matters" commentary — the commentary IS the product.
+- Fixed slots (intro note, main, shorts, footer) so readers learn where things live; consistent send day/time.
+- Write to one reader ("you", singular); a sender name that's a person ("Ayşe from Saglitz") typically beats a bare brand name.
+- Footer must contain: physical address, one-click unsubscribe, preference link, "why you're receiving this."
+
+## One Goal Per Email
+
+- **Every email has exactly one job and one primary CTA.** Multiple competing CTAs split clicks and blur the decision. Repeat the same CTA (top + bottom) rather than adding different ones.
+- CTA = bulletproof button (VML/table-based, not an image): background color renders without images, HTML text label, ≥ 44 px tall, action-verb copy ("Start free trial", not "Submit" or "Click here").
+- Secondary links (nav, social, footer) are fine but visually subordinate.
+- The inverted-pyramid layout: hook headline → supporting value → CTA button. A reader who sees only the first screen should be able to act.
+
+## Deliverability Basics Designers Must Respect
+
+- **Text-to-image ratio:** target ≥ 60:40 text:image by area; never send image-only emails. Filters and privacy-image-blocking both punish them.
+- **Authentication awareness:** SPF, DKIM, and DMARC are mandatory for bulk senders to Gmail/Yahoo/Outlook (enforced since 2024, tightening since). Designers don't configure them, but must know: no authentication = no inbox, and BIMI (logo in inbox) requires DMARC at enforcement + usually a VMC.
+- Additional bulk-sender requirements: one-click unsubscribe (RFC 8058 List-Unsubscribe header) honored within 2 days; spam complaint rate < 0.3% (aim < 0.1%).
+- **Spam triggers to avoid in design/copy:** ALL-CAPS subjects, excessive punctuation!!!, currency-symbol clusters, link shorteners (bit.ly), mismatched link text vs. href, forms and JavaScript (stripped anyway), single huge image.
+- **List hygiene:** double opt-in preferred; never purchased lists; suppress hard bounces immediately; sunset non-engagers (no opens/clicks 90–180 days) into a re-permission flow, then remove. Engaged-only sending raises inbox placement for everyone else.
+- Warm up new domains/IPs gradually (small volumes to engaged users first, scale over 2–4 weeks).
+
+## Accessibility in Email
+
+- Semantic structure: use real `<h1>–<h3>` (styled inline) and `<p>`, not styled `<td>` text soup. Set `role="presentation"` on layout tables so screen readers skip them.
+- `lang` attribute on the root; `dir="rtl"` when appropriate.
+- Contrast: 4.5:1 minimum for body text, 3:1 for large text — check both light AND dark-mode renderings.
+- Never convey meaning by color alone; underline links inside body copy.
+- Meaningful alt text (empty `alt=""` for decorative images); left-aligned body text (justified/centered paragraphs impair dyslexic readers); line-height ≥ 1.4.
+- Respect reduced motion: no strobing GIFs; keep animation subtle and non-essential.
+
+## Metrics Benchmarks (2026, cross-industry medians — verify per-industry)
+
+- Open rate: ~20–35% is normal (Apple MPP inflates; treat opens as directional only). Top decile senders exceed 40%.
+- Click-through rate: ~2–2.5% of delivered; click-to-open ~10–15%.
+- Automated/flow emails: ~5–6% CTR (≈3× campaigns); welcome emails 50%+ opens.
+- Unsubscribe: < 0.3% per send is healthy; spam complaints must stay < 0.1%.
+- Bounce: < 2%; conversion rate on promotional sends: ~0.05–0.2% of delivered (flows far higher).
+- Judge success on clicks, conversions, and revenue per email — not opens.
+
+## Progressive Enhancements (use, but never depend on)
+
+- **Interactive/AMP email:** carousels, accordions, in-email forms work in a minority of clients — ship only with a full static fallback, and never gate the CTA behind interaction.
+- **Gmail annotations / Promotions tab markup:** structured data (deal badge, image preview, expiry) can raise visibility in Gmail's Promotions tab — free lift for commerce senders.
+- **Litmus/Email on Acid–style rendering tests** before major templates; a template QA'd once per quarter drifts as clients update.
+- CSS you can rely on broadly: media queries, `padding`/`margin` on tables, web-safe font stacks, `background-color`. CSS you must fallback: flexbox/grid, `position`, web fonts, `border-radius` (cosmetic only), CSS animation.
+
+## Pre-Send QA Checklist (run every send)
+
+1. Subject ≤ 50 chars, key word first; preheader written and distinct from subject.
+2. From-name recognizable; reply-to monitored (no `noreply@` — it hurts trust and engagement signals).
+3. Renders correctly in: Apple Mail, Gmail web + app, Outlook classic desktop, Outlook mobile — light AND dark mode.
+4. Images off: message and CTA still comprehensible; all alt text present.
+5. One primary CTA; button is bulletproof HTML (not an image); all links tested and tracked (UTMs consistent).
+6. HTML < 100 KB (Gmail clipping); total image weight < ~1 MB.
+7. Live text for headline/offer/CTA; text:image ratio ≥ 60:40.
+8. Footer: physical address, working one-click unsubscribe, preference center.
+9. Personalization fallbacks set (`Hi {first_name|there}` — never "Hi ,").
+10. Correct segment/suppression lists applied; seed-list test email sent and read on a phone.
+11. Send time sanity: respect recipient timezone where the ESP supports it; avoid stacking sends < 48h apart to the same segment without cause.
+
+## Anti-Patterns
+
+- The all-image email (invisible with images blocked, spam-filter bait, inaccessible).
+- Designing only the light-mode version; shipping a dark logo that vanishes when Gmail inverts.
+- Multiple competing CTAs / trying to achieve three goals in one send.
+- "View in browser" as the preheader; subject lines that repeat as the preheader.
+- Discounting in the first abandonment email; training subscribers to wait for coupons.
+- Buying lists, skipping sunset policies, hiding the unsubscribe (raises spam complaints, which is worse than losing the subscriber).
+- Fake urgency ("last chance" weekly), fake "RE:" subjects — one-send wins that poison long-term sender reputation.
+- Ignoring Outlook because "it's old" — B2B audiences still skew heavily Outlook desktop.