Hello worldHello world

Master Playbook

The 25 most important lessons (best evidence first)

Built from kb/01 to kb/15 (9 Nov 2026). Every point cites the kb file and reel codes. Evidence tags: [data] numbers-backed, [demo] shown on screen, [anecdote] one person's story, [claim] unsupported. Most creators here sell a course, tool or service, so treat their "results" as marketing unless tagged [data].

Businesses in scope:

  • Ilene Joy (IJ): fine jewellery D2C, Shopify plus Instagram. Brand rules still apply: lowercase hashtags, no AI wording, never an AI label, approval mocks first.
  • Healthy Manushya (HM): India supplement D2C, COD heavy. Meta ads paused after about Rs 25k with about 0 sales.
  • Agencies: Zamn Studios, StudioHRTN, Goodform.
  • Own content: Tanish's personal reels.

Backed by numbers [data]

Backed by numbers [data]

Landing pages convert more when you remove things

  1. Landing pages convert more when you remove things. Seven named before/after tests (Yuppiechef, Imaginary Landscape, Career Point, Humana, Ring, Highrise) won by cutting nav, cutting form fields, cutting copy, one image, one obvious button. Lifts of 2x to 15x. Caveat: old single tests; removing nav suits single-goal pages, not browsing. kb/03 kb/15 DeLHuWuPiwj [data]

Most creatives lose

  1. Most creatives lose. Motion's study of 578,000+ creatives and $1.29B spend: only 3.7% to 8.8% become winners, lowest under $10K a month. Plan on 11 to 27 truly different concepts per winner. Test cheap simple formats first, polish only after a concept wins. kb/01 kb/02 DdoxpuXBQNF [data]

Shares are the strongest organic signal in these numbers

  1. Shares are the strongest organic signal in these numbers. The best reels had reshares near or above half of likes. Share drivers: frameworks, "send this to someone", consumer-protection explainers, identity content, contrarian rankings. kb/05 Dd7_w1wsdKf DcAkYEfyuPY DdCGW4wimBo DdtNBKpMLvQ DdjrOhAx9yE [data]

Comment-keyword gates inflate comments, not demand

  1. Comment-keyword gates inflate comments, not demand. Over 40 reels use them; comments often beat likes (160,837 comments on 2.29M plays). The count measures the CTA, not interest. Bots break at scale, so keep a bio-link fallback. kb/05 kb/07 kb/09 DeUUFJQu8cw De7JFJdo4KX DeABDB9I2D1 Dd6izTYiln1 [data]

Every visual claim needs a receipt on screen

  1. Every visual claim needs a receipt on screen. The biggest reel in the set (20.8M plays, about 1M reshares) paired every beat with code, a chart or UI. Proof-first openings (inbox, dashboard, order notification) recur in the strongest reels. kb/06 DecfvksvHiW Dd-rXZ8zqDd DcQ9S3EAaA_ [data]

Shown on screen [demo]

Shown on screen [demo]

Creative is the targeting now

  1. Creative is the targeting now. Meta's Andromeda step shortlists a few thousand ads per user by matching creative "vectors" before any auction. Change the reason to buy (desire, pain, persona, awareness stage), not the colour or one hook line. kb/01 kb/02 kb/11 Dd4fzTjR9rk DfXrim6xnyH DdpfvdHCDp1 [demo]

Clean Meta's defaults before every launch

  1. Clean Meta's defaults before every launch. Audience Network, Threads, Messenger off. Untick "Allow limited spending to excluded placements". Related media off. Creative Enhancements off. Multi-advertiser off for video. Re-check after publish and after duplicating. Judge placements on cost per real result, not CPM. kb/01 Dba8Pqvu-Y4 Db3L0XITUbY DecMToApLvz [demo]

In India, prepaid share moves with page trust plus money

  1. In India, prepaid share moves with page trust plus money. A clean, uniform PDP plus flat 10% prepaid off gave 44% prepaid on day one (n=9). A 15% discount on a low-trust, low-ticket store still left 80% COD. kb/03 DdoNputBhG1 [demo] Dc1rL8hSzS7 [anecdote]

High-ticket carts are worth a human

  1. High-ticket carts are worth a human. Alert on every abandoned checkout, call fast, close on a Shopify draft order. Claimed 60% of one store's revenue. Needs a phone number and consent. kb/03 kb/04 DcQ9S3EAaA_ [demo]

Respond to leads in seconds

  1. Respond to leads in seconds. A $1.4B remodeler called a test lead back in about 5 seconds; 60 seconds is the bar. One owner found instant calls hurt trust, so test timing. kb/03 kb/09 Dcjrk_HxEPR DdpPxHkIsuN [demo]

AI video needs a storyboard and art direction

  1. AI video needs a storyboard and art direction. Seven reels show one stack (Claude Code, Remotion, FFmpeg, Whisper, HyperFrames). Every serious workflow approves a storyboard before render. Default output looks the same for everyone ("Claude flop"); references, a brand sheet and banned effects fix it. kb/07 De8SPjEAuV0 DeUUFJQu8cw De2SEI7JP8r DemBKpbxSyd DeZKCA3PNQ7 [demo]

Process beats model for AI-built software

  1. Process beats model for AI-built software. A 20-item security checklist, a separate auditor run more than once, and CI tests decide quality. "One run finds about half." kb/11 kb/14 DeM46Iay29L De9bqWtoARc DfdtXLfAim5 DfJXMRLojO1 [demo/anecdote]

To get cited by ChatGPT

  1. To get cited by ChatGPT: question as the H1, answers in server-rendered HTML (Ctrl+U then Ctrl+F test), and a clear opinion. Google also says to fact-check all AI text, including titles, meta descriptions and alt text. kb/15 Ddl1taBAppw DeBJExFPKUF [demo]

Make the catalogue readable by AI shopping agents

  1. Make the catalogue readable by AI shopping agents. Plain category word in titles, short factual copy, GTIN, SKU, vendor, accurate price and stock, return policy in three lines. Check Shopify Agentic Storefronts. kb/03 kb/15 DeRoj0fymX7 Dd1HjInveV3 [demo]

A hook is a visual interrupt plus a real promise

  1. A hook is a visual interrupt plus a real promise. The gimmick earns the first second; the promise (question, topic, problem, outcome) earns the rest. Start mid-action, put the rules on screen, one contrast colour. kb/06 Dd8kCcCCGG3 DeSKm5HB3MF [claim/anecdote]

One person's story or opinion [anecdote/claim], but widely repeated

One person's story or opinion [anecdote/claim], but widely repeated

Know your money before you spend

  1. Know your money before you spend. Breakeven ROAS, RTO-adjusted results, cost per delivered order. Several "big result" reels (800x ROI, Rs 2.33 Cr) fall apart on this check. kb/01 kb/03 kb/13 Dc1D7FWyPGT DXd5NR2DvUZ DdZAwEftJSy DdjHwSZSpgy [claim]

Fix tracking before scaling

  1. Fix tracking before scaling. Run Conversions API server-side and optimise on events that mean money: Lead, MQL, SQL, Purchase for leads; confirmed or delivered order for India COD, not "order placed". kb/01 kb/03 Dd17XgUhDSo DcsX3MryYdy [claim]

Scaling makes numbers worse by design

  1. Scaling makes numbers worse by design. More budget reaches the next, lower-intent people. Judge marginal CPA, raise budgets in about 20% steps, consolidate campaigns, add new angles and products. kb/01 kb/13 DeRzDazODdV DdtatLDiS9g Dd83hWNjzaK [claim]

Put the value in the video, not the caption

  1. Put the value in the video, not the caption. "Read the caption" reels now annoy most viewers. End-load the payoff for watch time and design for one share type (to a friend, or to their own profile). kb/05 kb/08 Dd4oJ7gjqGX Dd4QXwYust3 [anecdote/claim]

Trial Reels are the safest test bed

  1. Trial Reels are the safest test bed. Test hooks on non-followers, then re-upload winners to main as a new post. Five creators describe versions of this; nobody has controlled data and duplicate flags are reported. kb/05 kb/08 DeSJg4Mkg42 DeFFzgiCQOe DeGVbSJtBfn DdwPZAku3Jx Dco8DB2oRRt [anecdote]

Followers are not customers

  1. Followers are not customers. Be known for one problem you have solved, repeat one message, show face, mistakes and results, then sell an owned offer. Small pages can earn (Rs 5.74 lakh in 76 days claimed, first deal under 5,000 followers). kb/08 DeERlGvM9LK Dd7_w1wsdKf DdT3NeiTKTx [claim/anecdote]

In saturated markets, lead with proof, guarantee and brand

  1. In saturated markets, lead with proof, guarantee and brand. Ladder: claim, unique mechanism, proof, guarantee, brand. Jewellery and Indian supplements are near the top. kb/10 kb/12 DaqtApmqK28 DdjRcDUybaT [anecdote]

Cart drop-off is the rational brain vetoing the impulse

  1. Cart drop-off is the rational brain vetoing the impulse. Fix shame ("is it for me?") with UGC, choice overload with 3 to 4 options, trust with reviews, 7-day returns and a founder video. Shorten impulse to payment. kb/03 DcfwuzkzK3w [claim]

Affluent buyers respond to specific ads, not "premium" words

  1. Affluent buyers respond to specific ads, not "premium" words. Selective placement, one specific benefit, a next step instead of a sale, a site that survives their research. kb/01 kb/10 DewY3OjCpRm [claim]

Price architecture beats discounts

  1. Price architecture beats discounts. A "show-off" decoy next to the best seller moves the anchor. A small paid first step ($3,000 before $50,000) lowers risk without discounting. Pick two acquisition channels you can still run in month twelve. kb/09 kb/10 DekuqC0IpMG DfAKJ9LCxql DdjybYQAbfu [anecdote/claim]

---


Consensus vs disagreement

Rules many reels agree on

The big disagreements, and which side has better evidence

Question Side A Side B Better evidence
Broad vs interest targeting Remove targeting, let Andromeda work (926 likes) Ddtkjn8ie0l Dc1D7FWyPGT Interest testing still wins in tier-2 India; a 4-campaign system keeps an interest CBO DdjHwSZSpgy Unresolved. No controlled data either way. Run a 7-day split for HM, judged on delivered orders. kb/01
Cut placements vs Advantage+ Turn off Audience Network, Messenger, Threads Db3L0XITUbY Dba8Pqvu-Y4 Meta UI says Advantage+ placements give 12% lower cost per result Neither shows lead quality by placement. Cut, then judge on cost per real result, not CPM. kb/01
How many creatives 150 to 750 ads a week DeFATG6umBR; 400 to 500 creatives Dd83hWNjzaK Small accounts have the lowest winner rate DdoxpuXBQNF Motion [data] favours fewer, truly distinct concepts at small budgets. kb/01 kb/02
Reposting and duplicates Post everything 5 times DeFFzgiCQOe; different hooks avoid detection DeSJg4Mkg42 Variants get flagged and deprioritised (commenters) Dco8DB2oRRt No Instagram docs cited by anyone. Repost outliers only, change the hook, never mirror product. kb/05
Speed to lead Call in seconds Dcjrk_HxEPR DdpPxHkIsuN One owner waits 10 minutes; instant felt creepy (724 likes) Majority view and the CRM demo favour fast. Test per business. kb/09
India RTO 60% fail, 80% COD Dc1rL8hSzS7 (second-hand, low-ticket) Operators report about 10% to 20% RTO (147 likes) Operators are closer to typical; measure your own by courier and pincode. kb/03 kb/14
Remove express checkout "Up to 40% AOV" without Apple Pay / Shop Pay DdCC7dXvIMz Shoppers will not buy without them Unsourced claim; keep express checkout unless an A/B test says otherwise. kb/03 kb/12
Gimmick hooks Gimmicks are not hooks Dd8kCcCCGG3 His own gimmick reel worked; top comment says "they are hooks" Both. Visual interrupt plus promise. kb/06
Caption-first reels Dead, viewers swipe Dd4oJ7gjqGX A 5-second b-roll plus long caption still hit 139k plays DdwPZAku3Jx The counter-case was keyword-driven. Value on screen wins. kb/05
AI-written scripts B to A tier with context Ddt4Tq0BGS_ Commenters say AI scripts "always fail" Use AI for research and illustrations, not first drafts. kb/06 kb/11
Brand consistency and logo Logo 8th, consistency 5th DdjrOhAx9yE Repeat exposure and distinctive assets drive recall Dd3qBGGyi0I; guidelines matter DbCzQQwRpgZ Do not over-invest early, but protect and repeat what already works. kb/12
Images for AI agents Fewer images, fewer popups DeRoj0fymX7 Model shots and rich galleries for humans DdCC7dXvIMz Humans still buy. Keep galleries; clean structured data instead. kb/03 kb/15
Dynamic creative Off, run individual ads Dba8Pqvu-Y4 Meta then concentrates spend on 3 to 4 ads DfXrim6xnyH One angle per ad set, 3 to 5 ads each, is the shared middle. kb/01

Action plan per business

3.1 Ilene Joy

Do this week

  1. Verify Meta tracking: Conversions API on the Shopify channel at "Maximum" data sharing, Event Match Quality, dedupe. Resolve the pixel mismatch noted in kb/01 before any paid push. Do not change the live pixel without a written plan. kb/01 Dd17XgUhDSo; kb/09 DdwDFt8Bsoj
  2. Confirm the Ilene Joy Business portfolio owns page, IG, pixel and catalogue, with two real admins and 2FA. Any agency gets partner access only. kb/14 DdwDFt8Bsoj
  3. Run the Ctrl+U then Ctrl+F test on PDPs, FAQ, shipping, returns and care pages. Check app-injected review and FAQ widgets first. kb/15 Ddl1taBAppw
  4. Time the Concierge and "Design your own" inquiry reply against the 60-second bar. kb/09 Dcjrk_HxEPR
  5. Look (read only) at Shopify Agentic Storefronts / Catalog settings and which products are eligible. Check region and default status in Shopify docs before changing anything. kb/03 kb/15 DeRoj0fymX7

Do this month

  1. Add a plain descriptor to every poetic title ("reverie gold evil eye charm necklace"), fill GTIN, SKU, material, karat, dimensions, and write returns in three lines. Test title changes on a few products first for SEO impact. kb/03 kb/15 DeRoj0fymX7 Dd1HjInveV3
  2. Abandoned high-value checkouts: alert the concierge, reply by WhatsApp or call within minutes (consented numbers only), close with a Shopify draft order or invoice. Extend "Save design" into a draft order. kb/03 kb/04 DcQ9S3EAaA_
  3. Start a Trial Reels test bed on the 2 to 3 best posts: 2 hooks (gift "for her", identity "my stack") times up to 4 edits (shorter, new grade, text, new cover). Never mirror lettering or asymmetric pieces. Re-upload winners to main as new posts. Log results in a lessons table. kb/05 DeSJg4Mkg42 DeGVbSJtBfn De-HXzJtpku
  4. Test four hooks on the same footage: price truth bomb, "if i were a dodgy jeweller" villain format (plating, karat stamps, carat total weight), explain-nothing hands placing a charm, a fill-in template. kb/06 Ddu65VYI0it DdCGW4wimBo DeXr6rKlB4G DcAkYEfyuPY
  5. Send a storyboard table as the approval mock for every reel. Captions stay below the neckline, off ears and wrists. kb/07 De8SPjEAuV0 DeUUFJQu8cw DeL07LCkYYr
  6. Profile: name field "Ilene Joy | fine jewelry", positioning line, proof, one link to the builder, three pins (builder demo, founder or designer story, customer favourites). Face content from Ilene or Rae, not Tanish. kb/08 DdrSAMWq-TP DeERlGvM9LK
  7. Build 3 to 5 question-headline guide pages with a clear Ilene Joy opinion ("solid gold vs vermeil for everyday charms"). Start a monthly baseline of what ChatGPT, Perplexity and AI Overviews say about the brand. kb/15 Ddl1taBAppw DeJF0mUva_N DdsLYSUPwGQ
  8. Run the 20-item security checklist as a prompt on the builder, share links, "email me my design" form, theme scripts and Shopify webhooks (HMAC). Add a Playwright E2E for builder to checkout and an LCP budget before each release. kb/11 kb/14 DeM46Iay29L DfdtXLfAim5
  9. Human fact-check every AI-assisted title, meta description, alt text and Journal fact (karat, sizes). kb/15 DeBJExFPKUF
  10. Write a one-page logo usage sheet (safe space, dark version, small icon, minimum sizes, misuse page) for press, stockists and creators. kb/12 DbCzQQwRpgZ

Test later

  1. A decoy hero piece (bespoke or high-karat statement) priced well above the core range, shown beside bestsellers. Also content. kb/10 DekuqC0IpMG
  2. A paid design consult credited to the final piece, before quoting bespoke work. kb/10 DfAKJ9LCxql
  3. A no-nav ad landing page (one hero, one CTA) through the existing gated-URL pattern. kb/03 kb/15 DeLHuWuPiwj DdpS97zoJBP
  4. A 6 to 10 second delayed email pop-up vs current. Matching the CTA colour to the selected metal swatch. "Frequently bought together" charms in cart. kb/03 kb/12 DdCC7dXvIMz
  5. A 30-second gift quiz that asks budget last and routes high budgets to Concierge. kb/03 DfdtwZ_PVxY
  6. A hidden-frame drop reel with a small perk (early access, extender), not 15% sitewide. Set up DM automation before posting. kb/05 Dd-rXZ8zqDd DdtNdBOFGHJ
  7. Seed hamsa, eye and North Star pendants with mindset or spirituality talking-head creators; ask for a tag and link. kb/08 kb/13 DdtNBKpMLvQ
  8. Paid ads on warm audiences (site visitors, engagers, ThruPlay) with affluent ad rules, Enhancements off so Meta does not alter metal tone. Check region and gender breakdowns weekly. kb/01 Dd3qBGGyi0I DewY3OjCpRm Dba8Pqvu-Y4 Ddtkjn8ie0l
  9. Plant buyer criteria only where true: "ask any jeweller: solid gold or plated? hallmarked? lifetime care?" kb/06 kb/10 DdkJX31x9LT

3.2 Healthy Manushya

Zero sales on about Rs 25k usually means offer, page, trust or tracking, not targeting. Fix those before any new spend. kb/13 Ddtkjn8ie0l DcfwuzkzK3w

Do this week (no spend)

  1. Audit the last Rs 25k: Breakdown by placement, hour, region, age, gender and frequency. Look for Audience Network, Messenger, WhatsApp and night spend. Write findings down. kb/01 Db3L0XITUbY DeWRfULIKr2 Ddtkjn8ie0l
  2. Build a per-order margin sheet: price minus COGS, ad cost per delivered order, RTO cost, packaging, shipping, gateway fees, GST. Set breakeven ROAS on delivered revenue. kb/03 kb/13 DXd5NR2DvUZ Dc1D7FWyPGT
  3. Measure delivered rate and RTO by courier and pincode, and COD share, from past orders. kb/14 Dc1rL8hSzS7
  4. Check pixel plus CAPI dedupe. Plan a confirmed-COD or delivered event via CAPI so Meta stops optimising for fake orders. kb/01 kb/03 Dd17XgUhDSo DcsX3MryYdy
  5. Ad Library teardown of 3 to 5 Indian supplement competitors: count active ads, find ones running 60+ days, note pain point, claim, offer and awareness stage. Any LLM can classify them. kb/11 kb/13 DdnGOKeRAn2 DXd5NR2DvUZ

Do this month

  1. Fix the PDP: trust strip (COD available, pay online and save, lab or FSSAI proof), review count near the title, 3 packs with the middle pre-selected and honest compare-at prices, founder video, 7-day return wording, sticky add-to-cart. No fake timers. kb/03 DduyHWQxw19 DcfwuzkzK3w
  2. Add a flat 10% prepaid discount in the announcement bar and checkout. Track prepaid % and RTO % daily. kb/03 DdoNputBhG1
  3. Add an explicit WhatsApp opt-in at checkout. Then run consented flows: COD confirmation within minutes with a UPI link, abandoned checkout, reorder reminder on day 20 to 25. kb/04 DeqkAyUR7uB DcQ9S3EAaA_
  4. Build 4 to 6 genuinely different concepts from the angle matrix (protecting family, energy, comfort or digestion, approval for skin and hair). Shoot a small 50-5-3: about 12 hooks times 3 meats times 1 to 2 CTAs, in Hinglish. The CTA says what happens next (COD, delivery days). kb/02 kb/06 DfXrim6xnyH DeFATG6umBR Ddu65VYI0it
  5. Compliance pass on every claim (ASCI, FSSAI, Meta health and personal-attributes policy). No weight numbers, no "you have X" framing. kb/02 Dezzkl-M3VR Ddu65VYI0it
  6. Restart organically: six tightly themed reels as Trial Reels, judged on shares and saves. Hero piece: "how to read any supplement label, even ours" in villain format. Verify FSSAI thresholds first. kb/05 kb/12 Dd_ZVdoI-iV DdCGW4wimBo
  7. Ad landing page: no nav, name, phone, pincode, one CTA ("order with COD"), proof above the fold. kb/15 DeLHuWuPiwj DdpS97zoJBP

Test later (when ads restart)

  1. One consolidated campaign, buyers excluded, settings checklist applied. One angle per ad set, 3 to 5 ads, judge after 2 to 3x target CPA. Expect about 1 winner per 25 concepts. kb/01 Dba8Pqvu-Y4 Dd83hWNjzaK DfXrim6xnyH DdoxpuXBQNF
  2. A 7-day broad vs interest split in tier-2 India, judged on delivered orders. kb/01 Ddtkjn8ie0l
  3. A "Reply to keywords" engagement ad that answers trust questions and sends a WhatsApp COD link. kb/04 DcbPON-zciJ
  4. A 6-month pack as the decoy next to the 3-month target. kb/10 DekuqC0IpMG
  5. An honest post-purchase "switch to subscribe and save" offer, prepaid only. Track take rate and first-renewal cancels. kb/04 Ddt306cRi2j
  6. Scale only after a winner holds: 20% steps, marginal CPA, ad scheduling instead of manual on/off. kb/13 DeRzDazODdV DeWRfULIKr2
  7. A founder "HM restart" quest series with real numbers. kb/05 kb/08 Dd7_w1wsdKf
  8. ChatGPT ads with a Rs 1k to 2k cap, low priority, judged on delivered orders. kb/09 DdjTSXCSriq

3.3 Agencies (Zamn Studios, StudioHRTN, Goodform)

Do this week

  1. Pick two acquisition channels you can still run in month twelve and cost them for 12 months. kb/09 kb/14 DdjybYQAbfu
  2. Add the asset-ownership rule to onboarding: client owns page, pixel, catalogue; agency gets partner access. kb/14 DdwDFt8Bsoj
  3. Export LinkedIn connections and score them against each agency's ICP in Claude (High, Maybe, Low). It is free and warm. kb/09 Ddi91WSz84Q
  4. Write a one-paragraph stance: no fake reviews or takedowns, no scraped audiences, no area profiling. It doubles as a sales point. kb/05 kb/11 DWnf25uCFzz Dd9GA7INc7H DdoiULSzAw_

Do this month

  1. Productise a paid first step (Rs 25k to 50k, credited to a retainer): Meta account audit (settings, CAPI events, placements, creative concept count, asset ownership), conversion audit (four vetoes, PDP ladder, Ctrl+U test), speed-to-lead test. kb/09 kb/01 kb/03 DfAKJ9LCxql Dba8Pqvu-Y4 DcsX3MryYdy DcfwuzkzK3w Dcjrk_HxEPR
  2. Build one own-lead funnel: direct-hook Meta ad to a Calendly "strategy session" page, pixel and CAPI on the booking. Test UAE, UK, US vs India. Track cost per booked call and per signed client. kb/09 DdZgna8Tzoh
  3. Sales process: answer-first decks with conclusion titles, a 15-minute pre-wire call per decision maker, roles decided in advance, the "you probably can" price script, value-first follow-ups with no "just checking in". kb/09 kb/13 DfPeqZATl-e DekqzGJxmrW DdU8uOPhGIK
  4. One creative brief template: angle (desire times awareness), hook family, meat format, five-part CTA, funnel stage. Price retainers on concept count, not ad count. kb/02 DfXrim6xnyH DeFATG6umBR Ddu65VYI0it
  5. One Notion template for all three: client OS, onboarding, win tracker, NPS. kb/13 Dcy173UgAVR
  6. Report lead-gen clients as cost per stage from real CRM data (clicks, leads, contacted, qualified, visit, close). kb/01 kb/09 DdZAwEftJSy

Test later

  1. A "front door" lead response product: one inbox, 60-second auto-reply with qualification, three real slots, revenue by source. kb/04 DdpPxHkIsuN
  2. AI-assisted reel editing with storyboard approval as the client gate, plus per-client brand sheets (the sellable part). kb/07 De8SPjEAuV0 De2SEI7JP8r DeZKCA3PNQ7
  3. A fixed-scope SEO plus AEO sprint with a before/after scorecard. No 3-day ranking promises. kb/15 DeI11acEwBI DdshNClk0t_
  4. Outbound with Harvey or a Maps pipeline only in a sandbox, after reading the code, with outbox approval, verified emails, SPF/DKIM/DMARC, 3-email max and opt-out. kb/09 DedgqT_TBjn Ddzd1fGqFZH Dd9dhDwt7gp
  5. AI agent builds scoped by the 8 layers; AI-built sites sold as "tested and secured" (20-item audit, CI, E2E). kb/11 DfJXMRLojO1 DeM46Iay29L DfdtXLfAim5
  6. Compare Skydo and Wise fee pages for foreign receipts and FIRA. kb/14 DdtoCqICWgU
  7. Logo handover checklist in every branding deliverable. kb/12 DbCzQQwRpgZ

3.4 Tanish's own content

Do this week

  1. Pick one solved problem and say it every time: "a solo founder running D2C brands with AI systems". Real artefacts are the credential. kb/08 DeERlGvM9LK Dd7_w1wsdKf
  2. Rebuild the profile as a sales page: keyword in name field, "I help [who] [get X] without [Y]", one proof line, DM keyword, one link, three pins. kb/08 DdrSAMWq-TP Dd6ZIHnBBwC
  3. Set up the "AI Video Editor" folder (raw, references, exports) with a brand sheet; the openfilm skill is already installed. kb/07 De2SEI7JP8r Dep0wuhvizo
  4. Start a hook swipe file tagged by family. kb/06 DcAkYEfyuPY Ddu65VYI0it

Do this month

  1. Launch a quest series with an open loop and real numbers: "I spent Rs 25k on Meta ads and got zero sales. Here is every number." Underdog, credible, and a reason to follow. kb/06 kb/08 Dd7_w1wsdKf DecfvksvHiW
  2. Run two hooks per reel through Trial Reels and log exit phrases with the AI retention loop. kb/05 kb/06 De-HXzJtpku
  3. Launch one keyword funnel with a real asset (the 20-item security prompt, Claude SEO project files, an angle-matrix sheet). kb/09 kb/11 DeM46Iay29L DeJF0mUva_N DfXrim6xnyH

Test later

  1. A free-beta price-discovery launch for any playbook. kb/08 kb/10 DdT3NeiTKTx
  2. Series formats: "Fact or Hype?" tool tests run in a sandbox, a ranking or tier list, "I timed 10 Indian brands' lead response". kb/05 kb/09 Dd6izTYiln1 DdjrOhAx9yE Dcjrk_HxEPR

Content system for Tanish's reels

Formats that fit (pick 4 and rotate)

  • Quest series with real numbers. "Day 1 of the HM restart." Open ending. kb/08 Dd7_w1wsdKf
  • Teardown with a visual receipt per beat. "I reverse-engineered X", every claim on screen. kb/06 DecfvksvHiW
  • Villain explainer. "If I were a [brand], here are 3 legal tricks, and how to beat me." kb/06 DdCGW4wimBo
  • Ranking or tier list. Shuffled reveal, one line of reasoning each. kb/06 DdjrOhAx9yE Ddt4Tq0BGS_
  • "Fact or Hype?" tool test, fail without it, work with it. kb/05 Dd6izTYiln1
  • One clip, N edits / slop vs directed. The reel is its own proof. kb/07 De7JFJdo4KX DeZKCA3PNQ7
  • Live packaging teardown. Weak cover, fixes, result. kb/02 De4rJ5SBSFk DfX5rchRb8i
  • Myth debunk with screenshots and a prop. kb/13 DXd5NR2DvUZ
  • Skit with a tool reveal. Panic line, blame joke, tool, proof panel. kb/06 De9bqWtoARc

Script skeleton

Hook (0 to 3 s, open a loop) then Explain (3 to 10 s, do not pop the climax) then Illustrate (10 to 30 s: example, story, maths, analogy) then Apply (1 to 3 takeaways). Ask Claude for 15 illustrations, keep 2 or 3. kb/06 DdjZgv9BFvf

Hook formulas

  • Four true hooks: promise, question, topic, problem. Add a visual interrupt in second one. kb/06 Dd8kCcCCGG3
  • Five ad families: problem agitation ("kya aapka bhi ___?"), contrarian truth, specific truth (provable number plus outcome), curiosity gap, truth bomb (price first, "yes it's expensive"). kb/06 Ddu65VYI0it
  • Fill-in templates with view proof: "the sound of X vs the sound of Y", "life is too short to ___", "life is hard, make it harder, start ___", "the answer to every problem is ___". kb/06 DcAkYEfyuPY
  • Breaking belief: state the belief, break it with proof. kb/06 Dd_ZVdoI-iV
  • Disclaimer first: "this is only if you want X, not Y." Builds trust. kb/06 Dd7_w1wsdKf
  • Explain nothing: 3 to 4 seconds of an unexplained act, then "you're watching ___". kb/06 DeXr6rKlB4G
  • Rules overlay: start mid-action, rules on screen, one contrast colour. kb/06 DeSKm5HB3MF
  • Delivery: rhythm plus tonal variation beats rushing or dragging. Record one hook four ways. kb/06 kb/07 DdFdFiavMJ0

CTA and keyword funnel setup

  1. Tie the keyword to a real asset (AUDIT, HOOKS, SEO). Do not gate a link that is already public. kb/05 kb/11 DeV2n53gLgf DeM46Iay29L
  2. Use two keywords: one low-friction resource, one offer (call or application). kb/09 DfAKJ9LCxql
  3. Set up automation before posting. Rotate reply texts. Add common misspellings. kb/05 DeABDB9I2D1 Dco8DB2oRRt
  4. Pin a bio-link fallback for when the bot overloads. kb/05 kb/07 DeUUFJQu8cw De2SEI7JP8r
  5. Avoid follow-gates; they often fail to deliver. kb/09 DdwPZAku3Jx DeLA2X4tapl
  6. A share-to-DM gate farms shares, a stronger signal than comments. Use it only for genuinely useful assets. kb/05 DeUK2NZS3Ig
  7. Write every CTA in five parts: what, why, when, how easy, what happens next. kb/02 DeFATG6umBR
  8. Judge the funnel on DMs answered, calls booked and sales, not comment counts. kb/05 Dd6izTYiln1

Posting and trial-reel cadence (synthesis from kb/05, kb/08, kb/09)

  • One shoot a week: 2 hooks with the same payoff, up to 4 edits each = up to 8 files. kb/05 DeSJg4Mkg42
  • Post 4 to 5 Trial Reels a week from those files. Re-upload the winner to main as a new post, not "share to main". kb/05 DeSJg4Mkg42 DeGVbSJtBfn
  • Repost outliers only at about +90 days with a new hook. kb/05 DeFFzgiCQOe
  • Use your own winner as the next template. kb/05 DdwPZAku3Jx
  • Stories: 3 a day (connection, value, proof) from the Claude calendar prompt. kb/05 Dd5MjjlKvr3
  • If content is a lead channel, the benchmark is about 15 pieces a week and about 6 months to the first call. Budget for that or pick another channel. kb/09 DdjybYQAbfu
  • Weekly review: paired-version retention, exit phrases, lessons table (lesson, evidence, counter, next test). kb/05 De-HXzJtpku

Editing stack

  • Agent: Claude Code desktop or OpenFilm. kb/07 De2SEI7JP8r Dep0wuhvizo
  • Tools: Remotion (captions, motion), FFmpeg (cuts, grade), Whisper (word-timed captions), HyperFrames (shot-level comments). kb/07 De8SPjEAuV0
  • Order: brand sheet from 3 to 5 references, storyboard table, render, one batch of notes, save the flow as your own skill. kb/07 De2SEI7JP8r DevtyU1kmHC
  • Sound: light SFX on cuts, licensed music, voice chain (high-pass about 100 Hz, light de-ess, mild compression, limiter). kb/07 De19MgmSv9N DeO-iyvFiCm
  • Framing: eyes on lens, titles at chest height, subtitles lower middle. kb/07 DeL07LCkYYr
  • Review cuts by hand. Cutting every pause over 0.1 s can sound robotic. kb/07 De2SEI7JP8r

Do-not-do list

Illegal or likely to breach privacy and platform rules

Illegal or likely to breach privacy and platform rules

Scraping competitor followers with emails and phones (Mailerfind) and uploading them as a custom audience

  • Scraping competitor followers with emails and phones (Mailerfind) and uploading them as a custom audience. kb/01 kb/09 Dd9GA7INc7H

WhatsApp messages to visitors who never opted in (Super Acquire style)

  • WhatsApp messages to visitors who never opted in (Super Acquire style). kb/04 DeqkAyUR7uB

Calling abandoned carts or sending cold email without consent and compliance (DND/TRAI, DPDP, CAN-SPAM, GDPR)

Excluding "slum areas" or any socioeconomic label from targeting

  • Excluding "slum areas" or any socioeconomic label from targeting. Exclude failing pincodes from your own courier or CRM data instead. kb/01 kb/11 DdoiULSzAw_

Fake countdowns, fake strike-through prices, false "3 spots left" or "closing soon"

  • Fake countdowns, fake strike-through prices, false "3 spots left" or "closing soon". India dark-pattern rules apply. kb/03 kb/06 kb/10 DduyHWQxw19 Dd5o0CboRhC

A discount labelled as an instant "refund" on a subscription upsell, with a timer

  • A discount labelled as an instant "refund" on a subscription upsell, with a timer. kb/04 Ddt306cRi2j

Fake reviews, "ambush" takedowns, AEO seeding with invented events or exact-match domain farms

  • Fake reviews, "ambush" takedowns, AEO seeding with invented events or exact-match domain farms. kb/05 kb/13 kb/15 DWnf25uCFzz DdsLYSUPwGQ

Hiding shipping times on the PDP

Ad policy and account risk

Ad policy and account risk

Fake news-headline or fake text-thread ads

  • Fake news-headline or fake text-thread ads. kb/02 DcelWe2yVu3

"3rd way" ads built on body shaming or sexual performance

  • "3rd way" ads built on body shaming or sexual performance. kb/02 Dezzkl-M3VR

Health claims like "15 kg in 30 days" or "protein is wrong"

  • Health claims like "15 kg in 30 days" or "protein is wrong". kb/02 kb/06 Ddu65VYI0it

"Revenue guaranteed in 60 days" style ad copy

  • "Revenue guaranteed in 60 days" style ad copy. kb/09 DdZgna8Tzoh

A 3-BM "firewall" built without a plan; same admins across many BMs can trigger restrictions

  • A 3-BM "firewall" built without a plan; same admins across many BMs can trigger restrictions. kb/14 DdwDFt8Bsoj

Turning campaigns on and off every night; use ad scheduling

  • Turning campaigns on and off every night; use ad scheduling. kb/01 kb/13 DeWRfULIKr2

Boosting posts for followers; use proper conversion campaigns

  • Boosting posts for followers; use proper conversion campaigns. kb/01 kb/05 Dd1MI7lhAYb

Security and unverified tools

Security and unverified tools

Installing skills or repos you have not read, especially from DM links or into ~/.claude/skills

  • Installing skills or repos you have not read, especially from DM links or git clone into ~/.claude/skills. Agent Reach drew "suspicious code" warnings. kb/05 kb/07 kb/11 Dd6izTYiln1 De7JFJdo4KX

Sending CRM, email or API keys to unverified models (Jev / TypeSafe)

  • Sending CRM, email or API keys to unverified models (Jev / TypeSafe). kb/04 kb/11 DdnGOKeRAn2

Pasting API keys into chat (Bland setup)

  • Pasting API keys into chat (Bland setup). kb/11 kb/14 Deb3ZCdgFjW

Running Harvey or any auto-sender without outbox approval and a token budget

  • Running Harvey or any auto-sender without outbox approval and a token budget. kb/09 kb/11 DedgqT_TBjn DezNz-3yeBc

Unverified, hype or misleading

Unverified, hype or misleading

Demo data shown as a client result ("800x ROI")

  • Demo data shown as a client result ("800x ROI"). kb/01 kb/09 DdZAwEftJSy

Buying .si domains on "AI renamed" urgency

  • Buying .si domains on "AI renamed" urgency. kb/12 kb/15 Dd6JQ1Fo7s0

"90 days to save your Shopify business" fear hooks

  • "90 days to save your Shopify business" fear hooks. kb/15 DeRoj0fymX7

"Reset the Stories algorithm" and "Edits gives a digital footprint boost"

The "female presenter cheat code"; it drew trolling and low-intent comments

  • The "female presenter cheat code"; it drew trolling and low-intent comments. kb/01 kb/02 DaAvfUkT_CD

150 to 750 ads a week on a small budget

Copying a commodity product from an "IndiaMART to Amazon" reel

  • Copying a commodity product from an "IndiaMART to Amazon" reel. kb/13 DXd5NR2DvUZ

Treating AI "consumer panels" as demand proof

  • Treating AI "consumer panels" as demand proof. kb/13 DeLA2X4tapl

Copying the decompile and reverse-engineering stunts without checking terms and law

  • Copying the decompile and reverse-engineering stunts without checking terms and law. kb/11 DeV2n53gLgf DecfvksvHiW

Brand-specific red lines (Ilene Joy)

Brand-specific red lines (Ilene Joy)

One-photo product-to-video for real jewellery shots; AI invents detail

  • One-photo product-to-video for real jewellery shots; AI invents detail. Mood and background only. kb/07 Ddtg-Ikhb6H

Mirroring reels with lettering or asymmetric pieces

  • Mirroring reels with lettering or asymmetric pieces. kb/05 DeFFzgiCQOe

Any "Human + Claude" or AI label on IJ posts

  • Any "Human + Claude" or AI label on IJ posts. kb/07 kb/12 DeEoDBASC5R

Downloading others' reels with yt-dlp or using trend audio without a licence on brand accounts

  • Downloading others' reels with yt-dlp or using trend audio without a licence on brand accounts. kb/05 kb/07 De8SPjEAuV0 DdwPZAku3Jx

Removing Apple Pay or Shop Pay without an A/B test

  • Removing Apple Pay or Shop Pay without an A/B test. kb/03 DdCC7dXvIMz

Countdown timers and free-gift stacking; they cheapen fine jewellery

  • Countdown timers and free-gift stacking; they cheapen fine jewellery. kb/03 DduyHWQxw19

---


6. Index

Index

File What's inside
kb/01-meta-ads.md Andromeda, Meta settings checklist, CAPI, account structure, scaling, bans and BM firewall, Reply to keywords, ChatGPT ads test.
kb/02-ad-creative.md Angle matrix, 50-5-3 modular ads, hook frameworks, funnel-stage formats, native formats, "3rd way", packaging rules, AI creative production.
kb/03-d2c-conversion.md Landing-page CRO tests, India COD and prepaid, four checkout vetoes, PDP ladder, abandoned-cart calls, catalogue-size layouts, AI-agent readiness.
kb/04-retention-crm.md Lead response systems, post-purchase upsell, WhatsApp consent, keyword and Story funnels, DM sales pillars, email deliverability, cohort reviews.
kb/05-organic-ig-growth.md Trial Reels systems, research methods, shares data table, keyword funnel table, Stories recipes, profile setup, formats catalogue, production basics.
kb/06-scripting-hooks.md What a hook is, hook template library, script skeletons, storytelling devices, delivery craft, copy and word choice, AI in scripting.
kb/07-video-production.md Claude Code editing stack, 4-step editor recipe, storyboard rule, effect skills, art direction, sound and voice chain, OpenFilm.
kb/08-personal-brand.md Followers vs customers, monetisation routes, profile as sales page, status and trust habits, follower formula, Trial Reels for personal accounts.
kb/09-lead-gen-sales.md Channel costs, speed to lead, Meta for quality leads, keyword and Story funnels, outbound agents, AI search inbound, sales scripts and negotiation.
kb/10-pricing-offers.md Market sophistication ladder, decoy pricing, paid first step, price objections, supplier positioning lines, affluent ad rules.
kb/11-ai-tools-automation.md AI tier list for creators, retention analytics loop, skills and repos with risks, security and CI for AI code, AI for ads, leads and SEO.
kb/12-branding-design.md Brand priority ranking, logo handover checklist, frequency rule, AI design tells, store design, honest label copy, thumbnail rules.
kb/13-strategy-founder.md Commodity-product trap, goal back-solving, Hormozi's 10 rules, founder communication, agency systems, pitching, scaling as strategy, teardowns.
kb/14-finance-ops-tools.md International payments for India, COD cash flow, Meta asset protection, channel cost sheet, supplier phrasing, open-source swaps, Oracle free tier.
kb/15-seo-ai-search.md Getting cited by ChatGPT, AEO steering risk, Google's AI-content rule, agentic commerce, local SEO tier list, layout by traffic, site security.

Supporting files kb/90 to kb/95 (hooks bank, tools and links, CTA funnels, formats, doubts and red flags, all links) exist in the same folder. They were not used as sources for this playbook.