Five audits · Growth Manager Deck, slides 10–38 Media buyer · copy chief · real buyer · CFO · deck designer

Five readers. Five different ways it breaks.

Each auditor read only slides 10 onward — the bets, the boards, the launch package, the ask — and audited their own layer: the mechanics, the craft, the reality, the math. Findings are tagged by severity. The consensus at the bottom is what more than one of them demanded independently.

Audit 01 · The mechanics M.K. — Meta media buyer $2–4M/yr on DTC hardware accounts. Reads the plan the way the auction will: learning phases, conversion volume, structure. Ignores the story entirely. “The plan reads numbers Monday. At this structure and an unstated budget, the account won’t have said anything by Monday.”
BlockerThe Monday read is arithmetic-impossible at any modest cap. Meta needs ~50 conversions per ad set per week to exit learning. At a $120 CAC that is $6,000/week minimum — and the deck never states the first-week budget. If the cap is below that, every Monday number is learning-phase noise, and the kill rules will execute ads on noise. Fix: state the number on the ask slide, and judge week one on 3-second view rate, hook rate and CPC — not CAC.
MajorThe 72-hour kill fires inside the learning phase. Killing and replacing ads every Friday resets learning weekly — the account never stabilises. Split the rule: creative metrics (hook rate, CTR) may kill at 72h; CAC may only kill after 50 conversions or 7 days, whichever comes first.
Major“Budget doubles weekly” re-enters learning every time. A doubled ad-set budget is treated by the auction as a new ad set. The scalable move is +20–30% every 2–3 days, or duplicating the winner into a CBO campaign. The tracker’s winner clause describes the one scaling method that reliably breaks winners.
Major“Retargeting only” for ad 16 is unspendable in week one. The retargeting pool off one new page’s traffic is a few hundred users — Meta cannot deliver against it. 16 isn’t a retargeting ad yet; it’s benched until the pool exists (roughly 10k page views). Say that, or the structure looks unexamined.
MinorOne ad set — except 04 is “gamer & tech audiences only.” That is a second ad set by definition, which splits the budget the launch package says is one. Either 04 waits, or the deck shows two ad sets and halves the conversion volume math accordingly.
MinorNo campaign type named. On a fresh pixel with six creatives, Advantage+ Shopping is the highest-probability week one; manual gives the cleaner angle read the tracker wants. It’s a real trade-off and the deck doesn’t make the call.
Audit 02 · The craft R.D. — direct-response copy chief Thirty years of control packages. Read only the boards — headline, primary text, caption, offer — and graded them as ads, not slides. “Three real ads and seven well-dressed variations. Ship 09, 02 and 01. Rewrite the rest around one category sentence.”
MajorThe category sentence appears in three different wordings. “A games console shaped like a puzzle,” “a console you twist, tilt and shake,” “the console you twist.” A new category gets named by one sentence repeated verbatim until strangers say it back. Pick one — I’d take “the game console you twist” — and put it word-for-word on all ten boards, the page, and the link cards.
MajorAll ten primary texts speak in one register — clever declarative, brand voice, under 20 words. Not one opens with “you.” Not one opens mid-story. Not one is long — and long primary text is a free test Meta rewards with cheap engaged clicks. Rewrite three of the ten: one second-person (“Your kid’s next screen shouldn’t scroll”), one story-first, one 150-word.
MajorNo ad ends with a command. The Shop-now button is chrome; the reader’s eye never gets an imperative in the copy itself. Every caption should end on one: “Watch the turn.” “Read the reviews.” “Get the bundle before December does.” The oldest rule in the trade, missing ten times.
Minor“One price, nothing stacks” is inside-baseball. The buyer never worried about stacking — we did. Say “$499. That’s the whole price.” Same honesty, buyer’s language.
MinorLink-card headlines are labels, not offers. “Full specification” is a filename. Every headline should carry the price or the promise: “See why it costs $399,” “The console that survives being dropped.” Labels waste the second-most-read line on the unit.
CreditWhere it’s genuinely good: 09 (“24 screens. Nothing scrolls.”) is a control candidate — it names the fear and owns the category in five words. 02’s price honesty is real direct response, not the imitation of it. And putting the worst comment in ad 05 is the kind of move that takes nerve — keep it.
Audit 03 · The reality check Jess, 41 — an actual parent, cold Two kids, 8 and 11. Shown the ten boards with zero setup — no deck, no explanation. Reported reactions, not critique. Her divergences from Dana (the simulated parent) are the finding. “I’d click the one that says nothing scrolls. I’d buy the one that says it survives being dropped.”
BlockerDurability never appears — and it was her first question. “Does it break when they drop it?” came up before price, before games, before screen time. A $399 object handed to an 8-year-old lives or dies on this, and none of the ten boards touches it. There is a durability-proof ad missing from the set — likely the highest-converting parent ad not yet written.
MajorShe only understood what it IS on two of ten boards — 20 (after its added what-is-it line) and 02. On 01 she guessed “a Rubik’s toy with screens? an app?” The copy chief’s one-category-sentence fix is confirmed from the other direction: comprehension, not craft.
MajorDana diverges from the real parent in three measurable ways. Dana opens comments twice; Jess opened none (“I don’t read ad comments, I ask the group chat”). Dana never asks about durability; it was Jess’s first question. Dana reacts to the STEM framing; Jess reported she scrolls past anything that says it teaches (“everything says that”). The scroll-test method survives — but its personas need calibrating against three to five more real buyers before the verdicts are trusted for edits.
Minor$399 is a “December decision.” Her words: “for us that’s a console decision.” Impulse framing didn’t land; the gift ad did. Supports the gift bet — and suggests the parent ads should sell the save-for-holiday plan, not the buy-now.
MinorTwo asks no ad answers: can siblings share it (a two-kid multiplayer angle — absent from all ten), and she read the anti-sell 16 as sarcasm from a brand she doesn’t know (“if I don’t know them, that tone is annoying”) — independently confirming the retargeting-only call.
Audit 04 · The math D.P. — operator CFO, hardware DTC Ignored every board. Audited the money spine only: where the ceiling comes from, what the offer does to it, what scaling does to cash. “I can’t approve a ceiling I can’t rebuild from margin. One bridge slide turns this from a request into an approval.”
Blocker$120 is asserted, never derived. At $399 with hardware COGS typically 40–50%, contribution before ads is roughly $180–220. A $120 CAC leaves $60–100 to cover payment fees, shipping subsidy, support, and returns — with a 30-day return window advertised on every board. That may hold; it may not. The deck needs the bridge, five lines: price → landed COGS → contribution → allowable CAC → the $120. Until it exists, every kill rule downstream hangs on an unaudited number.
MajorThe bundle ceiling is scaled by AOV, not margin. $150/$499 and $120/$399 are the same 30% ratio — but the bundle adds a dock and case at lower blended margin, so its true allowable CAC is likely below a proportional scale-up. Derive it from bundle contribution, not from the sticker.
Major“Doubles weekly” is exponential with no cash gate. Four doublings is 16× week-one spend inside a month. Hardware means inventory: can supply even absorb a winner? Add a monthly spend cap and one line on working capital — the scale rule needs a ceiling of its own.
MajorThe 15%-off offer and the CAC ceiling never meet. 15% of $399 is $60 — the discount alone consumes half the ceiling, and it sits on most boards while the tracker judges bets against the full $120. Either the ceiling is net-of-discount (then it’s effectively $60 and several kill rules fire immediately) or the discount comes off the boards. The two numbers cannot keep not knowing about each other.
MinorLTV is silently excluded. The games ecosystem presumably has attach revenue; excluding it from the ceiling is the conservative call — but say it: “ceiling assumes zero LTV; ecosystem revenue is upside only.” One sentence buys credibility.
Audit 05 · The deck itself S.L. — presentation designer Builds board and fundraising decks. Audited slides 10–38 as a presentation: pacing, hierarchy, legibility in a room, whether the structure serves the twenty-minute walkthrough it promises. “The system slides are disciplined. Then the creative section runs twenty-two slides on a two-beat loop, and the close is followed by an appendix that talks over it.”
BlockerTwenty-two slides of board–test–board–test is a metronome. By the fourth ad the audience knows the rhythm and stops reading; by the seventh the presenter is skipping. Present three ads in full (09, 02, 01 — the ones the other audits crowned), show the other seven as one contact-sheet slide, and move their scroll tests to the appendix. Twenty-two slides become six; the twenty-minute promise becomes true.
MajorThe Meta-unit chrome is illegible from two meters. 11–13px chrome text scaled to roughly 15px on a 1080p slide reads fine on a laptop and dies in a room. The chrome earns its place as proof of format — but the board’s headline should also appear at slide scale beside the unit, so the room reads the ad while the unit proves the format.
MajorThe appendix talks over the close. “The ask” is built to be the last thing seen — then the simulation prompt plays after it, and the room’s final image is a wall of mono text. Either end on the ask (appendix as a separate leave-behind file) or insert a black “Appendix” divider so the close visibly ends the argument.
MajorSection numbers and slide count have diverged. The kickers say 10, 11, 12, 13, 14 — but section 12 alone is twenty-four slides, so “12” appears on two dozen consecutive headers while the rail shows 12a through 12n. Number the sub-beats on the kicker itself (12.3 of 12.14) or drop the number from repeated headers and let the section name carry it.
MinorTwo competing divider languages. Week one is a red poster; the contents and the ask are black slides. Both read as “section break,” so the hierarchy between them is invisible. One color for chapter openers, the other reserved for the two moments that matter (contents, ask) — not interleaved.
CreditWhat’s working: the before/after convention is instantly legible and never explained twice; the kicker system on slides 01–09 is genuinely disciplined; the tracker’s thumbnail column solved its wall-of-text problem; and the red band running through week one, the verdict tags and the ask gives the deck one continuous voice. The bones are good — the creative section just needs an editor with a knife.
Consensus — demanded by more than one auditor, independently
1 · State the number. The media buyer needs it for learning-phase math, the CFO for the cash gate. “A capped week” convinces no one who can approve it.
2 · One category sentence, verbatim, everywhere. The copy chief demanded it for naming; the real parent proved it for comprehension — she understood two boards of ten.
3 · Reconcile the discount and the ceiling. The CFO caught it in the math; the media buyer’s kill rules execute against the wrong number until it’s fixed.
4 · Calibrate the simulated buyers against real ones. One cold read produced three divergences and the best missing ad (durability). Three more reads before the personas earn edit authority.
5 · Fix the scale rule. “Doubles weekly” fails both audits that touched it — it resets the auction’s learning and it outruns the cash. +20–30% steps under a monthly cap satisfies both.
What survived all four: the bet format itself (hypothesis, one variable, kill rule written first), ad 09 as the lead parent creative, the price-honesty ad, and thin tracking. Nobody attacked the spine — everyone attacked a number hanging off it.
Growth Plan Audit · five readers on slides 10–38 Auditors 1, 2 and 4 are role simulations; audit 3 is the one to replace with a real cold read first.