The operator playbook, updated for the two-day format and the $3m+ target. Seven topics spread deliberately from cheap-volume to premium-intent (the crypto vs generational wealth principle: the cheap end buys reach, the premium end buys the buyers you actually want). Run all seven simultaneously as Meta Instant Forms at $300 each, same webinar behind every one. That is $2,100 against the $1,800 originally approved: topic 7 is the addition, and if the line has to hold, it is the seventh cell or the sixth, not both. Every topic ships with a 60-second video hook in the OPC short-form structure, hook inside two seconds, one idea, CTA at the end.
The sibling of the webinar topic test, run against the free AI Search audit instead of the $7 seat. US-first, aimed at $1m+ businesses on $2k to $4k a month retainers over 12 months. Seven angles spread from broad reach to premium intent, same audit behind every one, $500 a cell and $3,500 in total. Lead form objective, not schedule: the closer has no cold ad experience, so the form and the VSL have to do the filtering before anyone gets on a call. The seventeen angles on the film day page are the creative bank. These seven are the test that decides which of them gets the budget.
Destination changed 3 September: the flight lands on sh.studiohawk.com, the VSL call funnel (VSL, five-question application, senior review, then the calendar), not the growth stepper. Round one (31 August to 2 September, seven cells, roughly $1,100 spent) ran on Advantage+ placements with no exclusions and an open Advantage+ audience, and the algorithm put 91% of each cell's budget into Audience Network in-app inventory. The people it found typed "Bills", "Nunya" and their own email address into the free-text question. The angle rankings from that round measure the wrong audience and are not carried forward. What is carried forward: the five cells below, run as ad variants under one correctly configured ad set, so the angle question gets answered on the right people while the flight produces bookings. Full readout on the growth dashboard.
?a= at 40. Angle 3's existing tag is exactly 40. Keep the existing tags on the four returning cells so the dashboard keeps their history in one row.Budget: at $3,500 that is about $700 a cell, enough to find one clear winner and one clear loser, not to rank the middle. If the flight comes in under that, let 57% More Per Lead sit it out; it is the one cell that is neither AI-search nor competitor-led and it will be the hardest to read at $700. Three of the five sit on the AI-search theme (proof, fear, doubt), which is deliberate: the test is now partly which emotional register carries that story, not five unrelated topics. The two things from round one worth keeping are the ad-set findings above and the form finding that the typed business-name step loses four in five people who reach it.
Same fact, two altitudes: the webinar cell sells it broad, the call cell names the one rival and pulls the citation gap live on the call.
Expected economics: maximum recognition, lowest CPL, widest junk net. The cheap anchor of the test: $1m+ answers showing up even here tells you the audience exists at volume. It does not tell you another topic will scale, that is what each topic's own cost per $1m+ answer is for. Destination fallback, only if Instant Forms are unavailable in the ad account: https://hawk-webinar-preview.pages.dev/?a=chatgpt
Who it attracts: the owner who is personally competitive and has a specific rival, and the marketing lead who has been asked by the CEO why a specific competitor is winning. Both are highly motivated. Expected economics: the strongest lever available to this funnel, because an abstract threat is easy to defer and a named rival is not, and because it is the one promise the free audit can deliver inside the call itself. The head-to-head that matters is against cell 1, which sells the same underlying fact in the abstract: if this wins on cost per qualified application, specificity beats scale and that finding carries into every cell written afterwards.
The survey's second-biggest fear. The call side runs it as the broad-reach anchor; the webinar side runs it mid-premium with status framing.
Expected economics: fear-of-missing-out with status framing pulls established owners. CPL will run higher than topics 1-3; the $1m+ share decides if it earns scale. Destination fallback, only if Instant Forms are unavailable in the ad account: https://hawk-webinar-preview.pages.dev/?a=leftbehind
Who it attracts: the marketing lead feeling pressure from above, and the owner who has heard AI search is the future and does not want to be late. Expected economics: the widest net in the test and the lowest cost per raw application, which also means the widest junk net. This is the cheap anchor: qualified answers showing up even here tells you the audience exists at volume. It does not tell you it will scale, that is what each cell's own cost per qualified application is for.
The rent argument. See also M2 in the deck bench below, which is this theme with the arithmetic doing the qualifying.
Expected economics: self-selects active spenders, the exact profile feeling the squeeze. A/B answers here are retainer-grade pipeline regardless of webinar conversion. Destination fallback, only if Instant Forms are unavailable in the ad account: https://hawk-webinar-preview.pages.dev/?a=adcost Resolved, 4 September: the 57% figure is superseded, not sourced. Harry's SOURCE manuscript cites the real one in its Notes: cost per lead on Google Ads rose 72% between 2022 and 2025, $40.74 to $70.11 (WordStream/LocaliQ Search Advertising Benchmarks), and paid clicks fell 68% on searches showing an AI Overview (Seer Interactive, September 2025). Both are printed with their sources on slide 20 of the live webinar deck. Swap every 57% on this page to 72% and give the dollar figures, which land harder anyway; the 68% was always safe and can stay. Rewritten scripts are on the call bank and the webinar bank.
Who it attracts: the owner watching paid margins compress, and the marketing lead who needs a diversification plan to present upward. Expected economics: the one angle in the source doc with no equivalent already written on the film day page, so it is the genuine addition rather than a re-cut. A cost statistic only stops people who pay that cost, which does the qualifying inside the creative. A and B answers on the second field are retainer-grade pipeline regardless of what happens on the call. Note: this cell reuses the 57% figure, now superseded by the sourced 72% ($40.74 to $70.11, WordStream 2022 to 2025); see the resolved note on W2 and swap before running.
The highest-volume pain in the survey. Admitting confusion is the hook, so both cells grant permission rather than shame.
Expected economics: the control, our highest-volume enquiry topic. Watch for A answers paired with $1m+ revenue on retargeting: those are call-funnel candidates, not webinar buyers. Destination fallback, only if Instant Forms are unavailable in the ad account: https://hawk-webinar-preview.pages.dev/?a=howto
Who it attracts: the marketing lead who has read everything and decided nothing, and the owner who knows this matters and has no way in. Expected economics: the highest-volume pain in the survey at 15 mentions, and until now the call funnel had no angle against it at all, which is the gap this fills. It is also the one angle where admitting confusion is the hook, so the creative has to grant permission rather than shame. Watch the failure mode carefully: this pain is strongest below the ICP floor, because the people most lost are usually the smallest, so the revenue answer is doing more work here than in any other cell. If the qualified share is poor but volume is high, do not read that as the angle failing, read it as the angle needing a harder revenue gate in the creative itself.
The only proof-led theme, and historically the strongest premium creative on the webinar side. Without it the test never asks whether proof outsells pain. The vertical split below argues the same territory audience by audience: the 6sense study for B2B, the consideration set for ecommerce, the referral argument for professional services.
Expected economics: proof-led, pulls people who think in outcomes not tactics. Historically our strongest premium creative; expect the best $1m+ share per dollar. Destination fallback, only if Instant Forms are unavailable in the ad account: https://hawk-webinar-preview.pages.dev/?a=samsung
Who it attracts: people who think in outcomes rather than tactics, which skews to owners and senior marketers. Expected economics: a restore, not a new idea. The source doc has no proof-led angle at all, every one of its eight is problem or fear led, so without this cell the test never asks whether proof outsells pain for this buyer. It is also the story the VSL already tells in beat 2, so the ad and the video agree by construction. Historically the strongest premium creative on the webinar side, and this second field is the ICP criterion the other cells do not measure.
The recognition pitch translated vertical by vertical, each with the filter built into the copy: the 6sense research for B2B, the consideration set for ecommerce, the referral argument for professional services. Same rule as the deck bench: creative variants inside W5 / C5 (or W6 for ecommerce), not new budget cells, and the unscored second fields read the vertical mix of the account.
Source: the shortlist-maths slide, day one of the deck (6sense, approximately 4,000 B2B buyers). This is Samsung's theme argued with evidence instead of anecdote: 5 says "it happened to us", 5B says "here is the research showing it happens to everyone", and the head-to-head between them is a clean read on whether this audience trusts a story or a study. The creative is also a B2B filter by construction, which no other cell has: the numbers mean nothing to an ecommerce owner, and the deck's own note applies here, do not stretch them to ecommerce, the machine writes the consideration set before anyone opens a tab is the one permitted line. Expect lower volume and a higher qualified share, which suits the call side better than the webinar side. Recommendation: same rule as the deck bench, run it as a creative variant inside W5 / C5 rather than a new budget cell, and if the account skews B2B on the second field, this variant inherits the Samsung budget in round two.
The ecommerce translation of the recognition pitch, anchored on the one proof asset built for it: Sylvane, which W6 already carries as the premium anchor, so this variant and W6 share a fact and must not run against each other in the same stack. The deck's own guardrail applies: the 6sense numbers are B2B research and stay out of this creative entirely, the consideration-set line is the one permitted claim. The second field reads borrowed-channel dependence, and a B answer is the strongest fit in the test for the owned-channel argument: a store that lives on Amazon is one policy change from zero. Watch the failure mode, which is AOV: plenty of stores clear the revenue floor on thin margin and cannot carry a retainer, so on the call side the revenue gate does less work here than anywhere else. Skews webinar: the $7 seat and the toolkit fit a store owner's buying pattern better than a strategy call does.
The professional-services translation, and the natural home ground of the whole pitch: StudioHawk is a professional services firm, so the Samsung story is first-person proof here in a way no other vertical gets, the ad and the firm are the same case study. "The referral you cannot ask for" is the frame, because a referral-built firm already believes in exactly the mechanism AI recommendation now automates, and the second field reads that dependence directly: an A answer is a firm whose entire growth engine is the thing the machine just started doing at scale. Best retainer fit in the trio, so it skews to the call side. Watch the failure mode, which is the solo practitioner flood: this pain is sharpest below the revenue floor, sole traders and two-partner firms will fill the form, so the revenue gate does the most work here and the creative should say "firm", never "freelancer". Head-to-head worth reading: V3 against C2, the named-rival cell, because in professional services the rival is usually one specific firm and envy may outpull the referral frame.
Webinar only. The purest $3m+ signal in the test; on the call side the proof job is done by Samsung.
Expected economics: the generational-wealth end of the spectrum. Highest CPL in the test by design, and the purest $3m+ signal: compare it on the one metric that decides everything, cost per $1m+ answer, meaning A or B only. If it beats every other topic there, premium-intent creative wins the account and the call funnel gets its own campaign next. Destination fallback, only if Instant Forms are unavailable in the ad account: https://hawk-webinar-preview.pages.dev/?a=sylvane
A deliberate cross-funnel experiment: the webinar sells three skills for $7, the call gives Structure Audit away for an email. Whichever side wins tells you where the toolkit belongs in the funnel.
Who it attracts: the person who buys instruments rather than outcomes, and the marketing lead who needs something tangible to show for a training spend. Expected economics: the only cell that leads with what they get rather than what they are afraid of, which makes it the cleanest read on whether this audience buys tools or buys relief. It is also the closest cell to the free skill funnel, so treat the pair as a deliberate experiment: this one sells three skills for $7, that one gives Structure Audit away for an email. If the free version pulls better qualified traffic at a lower cost, the toolkit belongs at the top of the funnel and not behind the ticket. Watch the failure mode, which is that tool-led creative attracts hobbyists who install everything and buy nothing.
Who it attracts: the buyer who trusts instruments more than claims, and the marketing lead who already has data and cannot get a decision out of it. Expected economics: the only angle here that gives something away before asking, which should pull a lower cost per application and a higher junk rate, so it lives or dies on the qualified share. A answers on the second field are the ideal buyer for this funnel: they have already done the diagnostic and stalled at the interpretation, which is precisely what the call sells. Run it as a deliberate pair with the same angle on the webinar test, which sells three of these skills for $7 while this one gives Structure Audit away: whichever side wins tells you where the toolkit belongs in the funnel.
Bench on the webinar side, live cell on the call side. The biggest real segment in the market: the people already paying an agency.
Lifts its hook and sub-topics from the VSL angles doc, retargeted from "book a call" to the two-day seat, and it ships with its sub-topic bank already written, so if it wins it has its next round of variations ready without a second writing pass. Adding it takes the test to eight cells and $2,400. The honest swap, if the line holds nearer $1,800, is to run it in place of whichever cell finishes last on cost per $1m+ answer in round one, rather than diluting spend across eight.
Who it attracts: the marketing manager who needs to look good internally and hold someone accountable to the CEO, and the owner who has been quietly suspicious for months. Expected economics: the biggest real segment in the market and the one gap in the current six, none of which speak to somebody who already has an agency. This ran in the original five as the DIY audit cell and was dropped when the format moved to two days, so treat it as a restore rather than a new idea. The unscored answer is the prize: A and B paired with $1m+ are call-funnel pipeline, not webinar buyers, and should route to the free audit rather than the seat. Watch the failure mode, which is that anti-agency framing pulls price-sensitive DIYers under $250k and the cell wins on raw CPL while losing on the only metric that counts.
Who it attracts: the marketing manager who needs to look good internally and hold someone accountable to the CEO, and the owner who has been quietly suspicious for months. Expected economics: this merges the ripped-off angle and the revenue-not-rankings angle, which are one insight said twice, that the thing your agency reports is not the thing you care about. It speaks to the biggest real segment, the people who already have an agency, and its second field is the closest thing in the test to a direct ICP read. Watch the failure mode: anti-agency framing pulls price-sensitive DIYers below $250k, and the cell wins on raw CPL while losing on the metric that counts.
Same buyer as cell 3, the business already paying an agency, with the accusation removed. "Ripped off" pulls the suspicious, and the suspicious skew to the $2k budgets who resent paying anyone. "Talking about AI SEO yet" keeps the person who values their agency and is quietly worried it has gone stale, which is the switch candidate worth having. The keystone objection is status quo ("what I have is fine"), and the Febreze move applies: they do not know they have a gap because nobody has raised it, so the hook is education dressed as a question, the proof handles trust, and the offer makes asking cost nothing.
Who it attracts: the owner or marketing lead who already pays for SEO, thinks it is probably fine, and has a small doubt they have not said out loud. Expected economics: the largest addressable segment for an agency switch, reached without making them defend their current supplier. The second lead-form field is the point of the cell: it sorts "agency has a plan" from "mentioned it, no plan" from "not once", which is the segmentation no other cell produces, and B and C answers paired with $1m+ revenue are the call-funnel's best prospects. Built on the OPC ad body (hook, problem, solution, proof, reminder, CTA) and the hook / proof / how caption. The reminder line, "if they are ahead of it we will tell you", reframes the price of asking to zero, which is what collapses the status-quo objection. Voice note for the film day: no contractions on the graphic or primary text, same as the rest of the bank; the 15-45s beat carries the Samsung proof so the ad and the VSL agree by construction. Round-one note: the original cell 3 had the best fit score of the seven and the worst efficiency, 38% of its respondents at $100k+ and 2.9% of its clicks ever qualifying; thin data on the wrong audience, but the direction, right people found expensively, is what this rewrite is trying to keep while fixing.
Lifted 18 August 2026 from the day-one exercises in the deck (the Mirror, the rent ledger). Both are creative variants for existing cells, not new budget cells: the test stays readable at 8 webinar + 7 call.
Source: the Mirror exercise, day one of the deck (slide: ask the machine about your own category). The ad performs the audit before the click, which no other cell in the bank does: a MISSING answer arrives already convinced, and the unscored second field is a self-scored mirror card, the closest thing paid traffic can get to the webinar's peak moment. Recommendation: run it as a creative variant inside W1 and C2 rather than as a new budget cell, because it sells the same recognition promise and a ninth cell muddies the cost-per-qualified-answer read. Watch the failure mode, which is the reassurance leak: people who run the prompt and get NAMED feel fine and never click, so the creative must plant "named is not the finish line, read where in the sentence you were" before the CTA.
Source: the rent ledger and rent benchmarks, day one of the deck (lines A to D, "circle line D"). This is the rising-ad-costs theme with the arithmetic doing the qualifying: a $40k-a-month spender circles $480,000 and feels it in their chest, while someone spending nothing has no number and self-excludes, which is exactly the filter W2 and C4 pay a second form field to approximate. "The rent went up. The street got quieter. Same landlord" is the best closing card available to any video in this bank. Recommendation: round-two replacement creative for whichever of W2 / C4 underperforms, not an additional cell. The deck's other exercises earn their keep elsewhere: rented-borrowed-owned and the QBQ prompt are nurture-email material (they teach rather than stop a scroll), and the shortlist maths (60% / 80% / 95%) belongs in the VSL, not a thumb-stop. Note: this cell reuses the 57% figure, now superseded by the sourced 72% ($40.74 to $70.11, WordStream 2022 to 2025); see the resolved note on W2 and swap before running.