Alaskan King Crab Co. · Creative Demand Model
What we launched, and what Q4 asks for
January to August is what the account actually launched. September to December is the recommendation. Hover any month for its numbers and the reason behind them.
Ads launched = distinct Meta ad names whose first recorded spend falls in that month; media is Meta plus Google. August is partial, through the 19th. The naming convention changed in April 2026 — one ad name per asset variant instead of one per concept — so months before and after that line are not directly comparable, which is why Section 2 argues the volume from spend and calendar rather than from the trend.
Why we should make this many ad creatives
Media rises, and creative scales with it
Q4 is when spending power improves
The calendar is dense and the windows are short
The funnel is narrow, so volume is the input
Launch volume went from 22 ads a month across January–March to 193 across April–August — nearly nine times — and the count of ads clearing $2,000 did not move: 55 in 2024, 53 in 2025, 49 across the first eight months of 2026. More shots find more winners in absolute terms, but the conversion rate is the ceiling. Grade the quarter on funded ads, not on ads launched.
June already proved this works at peak
Father’s Day is the closest rehearsal for Q4 the calendar offers — a compressed gifting window with a hard shipping cutoff. The brand attacked it with volume and the efficiency went up.
What June tells us about November and December
Where the returns actually came from
Twelve dated windows, each sized to its own budget
Every dated moment confirmed through year end, plus the three shipping cutoffs that behave like promotions without a discount attached, and the giveaway split across the two months it runs. Each window is sized from its share of that month’s promotional budget.
When each set has to be live
Hatched is the build runway, solid is the window itself. Lead times are set by how much of the mechanic is already proven: six weeks for a flagship built from scratch, two to three weeks for a re-cut of something that ran this year, one to two for a dated urgency asset.
Thanksgiving repeats the July 4th offer, the December Holiday promo repeats Memorial Day, and NYE repeats the Father’s Day free-Tomahawk mechanic. That lowers the risk in this plan materially: July 4th is the highest-returning seasonal tag in the account at 5.93x and produced the largest ad of the year — the “your guests will never know” hosting angle at $25,237 and 6.19x. Thanksgiving should start as a re-cut of that winner. Father’s Day returned 2.09x, so NYE inherits a weaker record and should be funded accordingly. Black Friday is the one set with no 2026 read behind it.
What to build, month by month
The headline unit is a launched ad. Concepts are the production sub-unit — the account runs 1.77 cuts per concept, so the concept column is what gets briefed.
| Month | Media spend | Evergreen | Moment build | Moment top-up | Total ads | vs run-rate | Concepts | Video | Image | $ / ad | Expect funded | Expect scaled |
|---|
Against 2025, measured on ads that earned budget
Raw launch counts cannot be compared across April 2026. The comparable series is how many ads earned budget — those are the columns that matter.
| Month | 2025 launched | 2025 cleared $500 | 2025 cleared $2k | 2025 media | 2026 plan — ads | 2026 media | Spend YoY | Expected funded | Funded YoY | Expected scaled | Scaled YoY |
|---|
Expected counts apply the account’s own measured conversion rates — 11.6% of launched ads clear $500 and 4.7% clear $2,000, from the May–July 2026 launch cohorts — to the planned volume. Recent cohorts are biased low: an ad launched in July has had less time to clear $2,000 than one launched in May.
What these ads should actually be
Video stays the scale engine at 87.5% of spend and 3.02x. Image wins more often at a third of the cost, and Q4 is an offer-led quarter — which is what statics do best.
| Month | Video % | Video ads | Image % | Image ads | Carousel % | Why |
|---|
- Stop briefing carousels. 23 creative groups in 90 days drew $514 between them — less than one funded ad is worth — and not one has ever cleared $500.
- Push statics into the promo windows. Among funded creative, 52% of images beat the blended 3.05x against 42% of videos, at a third of the CPM.
- Cut Greenscreen and POV. 8.7% of spend at 2.06x and 1.5% at 1.51x, both well below blend.
The signal library
Measured across the trailing 90 days in Motion against a blended account ROAS of 3.05x. Share is share of Meta spend. Lean in underfunded relative to return · Hold correctly funded · Cut overfunded · Test promising but too small to call.
Hook tactic
Messaging angle
Intended audience
Visual format
The highest-returning signal in every category points the same way: Offer Only hooks (6.72x), Price Promotion angles (5.44x), Offer-First Banner as the visual form (3.48x), Social Hosts and Celebrators as the audience (3.46x and 3.57x), with Urgency hooks at 4.91x just behind Offer Only. A quarter built around hosting a table, a dated deadline and a stated price is not a compromise on brand — it is the intersection of everything this account already proves works, and it has been chronically underfunded against Relatability hooks, which take 13.6% of spend and return 2.50x.
Dated decisions, and the assumptions to watch
The plan is a starting position, not a commitment. Three of these are decisions with dates attached; the rest are the assumptions to watch and what to do when one moves.
- Read November’s efficiency against June before December’s build is locked. June is the only clean read the year offers on what this brand does inside a peak: media rose 38.0% and aMER rose 15.9% with it, from 2.45x to 2.84x. November asks for a smaller lift on more money — 3.31x to 3.70x, +12.0%, on 40% more media. That is the number to track live. If cumulative November aMER is tracking at or above 3.70x through the Thanksgiving window, December’s build stands as planned. If it is running short of the June-style improvement, the December moment volume comes down with it rather than being produced against media that will not be there.
- Use the three November windows as sequential checkpoints, and recalibrate December from the reads. Thanksgiving (Nov 15–18) is the first live test of promo creative at Q4 demand, the shipping cutoff (Nov 24) tests urgency, and BFCM (Nov 27–30) is the full-scale read. Take a decision on 1 December with three weeks of December build still ahead: if the first two windows clear their efficiency plan, hold December at 362 ads; if they miss by more than roughly 10%, pull toward the Lean line at 313 and keep the media where it is. December’s Holiday promo does not open until the 13th, so there is genuine room to re-cut.
- Bank the sale-moment learnings now and point all of them at Black Friday. Every 2026 promo window — Memorial Day, Father’s Day, July 4th, Free Ship July, and the September and October flights as they run — should be logged with its winning hooks, formats, offers and creators before BFCM briefing opens. The evidence is already unambiguous about where that leads: July 4th is the highest-returning seasonal tag in the account at 5.93x, Price Promotion is the best messaging angle at 5.44x, and Offer Only is the best hook tactic at 6.72x. BFCM is the one Q4 window with no 2026 read of its own, so it should inherit everything the other windows learned rather than starting from a blank brief.
- Lead Black Friday with the biggest offer the business is willing to put up, and make the offer itself the creative. Promo-tagged creative returns 3.65x against 2.64x for No Offer — a 38% premium the account only collects when there is a real offer to promote. The BFCM set should be weighted hard toward clear, high-contrast statics that land the value in a single read: the discount, the product, the deadline, nothing competing with them. Offer-First Banner runs 3.48x and image creative wins 52% of the time among funded ads against video’s 42%, at a third of the CPM. Narrative UGC is the wrong instrument for four days of peak intent.
- Build the hype window before the offer opens. The 98 ads behind BFCM are not all for the four days themselves. A meaningful share should run in the two weeks prior as teaser and list-building creative — announcing that something is coming, capturing the email and SMS file, and warming the audiences the offer will convert. The lead-gen giveaway running through 30 October feeds the same file. Deadline creative converts an audience that already knows the offer is coming.
- Confirm or drop the October retention promo by 21 September. It is modelled at 11 ads on the assumption it stays CRM-led with retargeting-only paid support. If it becomes a prospecting offer it needs roughly triple that, and the September build load rises with it — so the decision has to land before the September slate closes.
- Grade the quarter on funded ads, not ads launched. Launch volume went from 22 a month in January–March to 193 in April–August and the count clearing $2,000 did not move. Launch count is an input the team can always hit; funded count is the outcome that pays for the quarter. It is also the only year-over-year measure that survives the April naming-convention change, which makes raw counts before and after that line incomparable.
- Revisit the per-ad rates if the engine changes. $557 per moment ad and $744 per evergreen ad come from May, June and July 2026 — the only three months where the current testing engine and the current naming convention were both in force. Under the pre-April regime the account absorbed roughly $5,642 of media per launched ad, eight times these rates. If budget starts concentrating on fewer creatives again, these volumes are too high and the model needs re-cutting.
- Scope Google separately before committing the team. Volume here is driven by total media — $691,844 across Meta and Google, exactly as the Full-Year Forecast states it — on the basis that winning Meta creative feeds PMax and Shopping asset groups, which is how the account runs today. Ad-level history exists only for Meta, so the counts are expressed in Meta ad names. If Google needs bespoke assets rather than re-cuts, that is capacity on top of these numbers.