Alice is winning attention. It's losing the sale.

The product works. The math doesn't, yet.

The ads grab attention and people click, so this isn't a brand problem. The account is under a dollar for three fixable reasons, and Alice already proved it can clear a dollar return (it did in January). Fix these three things and the same spend pays back more.

The prize. Get the same $800K of media back over a dollar, the level Alice already hit in January, and the $651K it earns today becomes about $858K. That's roughly $200K more every quarter, about $825K a year, with no extra spend. We can start in week one.

Spend tripled and the account climbed above a dollar return. Then ROAS rolled back off.

From roughly $105K a month last summer to a $382K peak in December, Alice scaled Meta about 3.6x and pushed ROAS from 0.6 to above 1.0 in January and February, the only two months above a dollar all year. Then March and April gave it back, 0.86 then 0.71, even as spend pulled in. Trailing 90 days sits at 0.80x. The media scaled. The library aged into the scale.

A note on "above a dollar." 1.0x means Meta-reported revenue equal to spend, not the profit line. Assuming a 60% contribution margin (our working estimate for premium DTC functional food, to be confirmed against Alice's actual margin), real first-order breakeven sits around 1.67x ROAS. Which means even the Jan peak (1.07x) was running on LTV slack: a 37% subscription rate and recurring charges making the math close over the buyer's life, not on the first order. The 0.80x today doesn't leave enough room for that slack to do the work. The path is two-stage. Stage one is the prize this deck sizes: back to the 1.06x Alice already proved, roughly $200K more revenue a quarter. Stage two is pushing basket and subscription mix until first-order ROAS climbs toward 1.67x and the economics aren't carried by LTV alone.

One note on the number. We count a sale only when someone clicks the ad and buys within 7 days (the strict standard, and the one Motion uses): that's the 0.80x in this audit. Meta's own dashboard also takes credit when someone merely sees the ad and buys later, which flatters the number up to about 0.92x. We use the stricter read. Either way, the account is under a dollar. One housekeeping item: the ad sets are set to a mix of these windows today, worth standardizing.

Meta month-over-month.

Source: Motion / Meta account pull, May 28, 2026. May row is partial month (1–28). ROAS = purchase value ÷ spend, account-level. The 90-day and trailing-year headline totals are exact-window pulls, so they won't sum to the calendar-month rows exactly.

What drove the year, and what didn't. This was a spend ramp, not a holiday push. The library is overwhelmingly everyday rather than seasonal, and account-wide most spend runs without a discount. One important nuance: the Jan-Feb peak leaned hard on the Discovery Set's 30% off offer language in the top performers, so the proven peak shape was specifically "Discovery Set + 30% off," not Discovery Set alone. April-May has pulled away from that offer-led look, and ROAS slipped with it. The standing value offer (subscribe-and-save and free shipping) still pulls the best return at 0.98x against 0.76x no-offer, which is value, not markdown. Two open questions for the next pull: how much of the Jan-Feb climb was offer-driven vs creative, and whether the discount-led buyers stuck around or bought once and left.

The basket got bigger, not smaller. Average order value (AOV, the size of the typical basket) climbed all year, from roughly $72 last summer to $94.60 in the last 90 days. The problem isn't what people spend. It's what it costs to get them: the price to land one order (CPA, cost per acquisition) is now $117.55, about 20% above the trailing-year average of $97.77.

Alice is paying $118 to land a $95 order.

Over the last 90 days, 16.3 million people saw Alice's ads and 2.71% clicked through (the click-through rate, or CTR, well above the commodity field). So attention isn't the problem. The leak is the gap between what it costs to win an order and what that first order is worth. Close it, and the account clears a dollar without a single new click.

Read it top to bottom: cheap attention (CPM is the cost per 1,000 views), strong clicks, then the break. Every order costs more to win ($117.55) than it's worth on the first purchase ($94.60). That $23 gap, on every order, is the leak at the heart of it. Click metric definition (all clicks vs. link clicks) is one of the small alignment items for the next pull.

The basket is the lever. The set already proves it.

AOV derived from ROAS × spend ÷ purchases, last 90 days. The set clears its acquisition cost; the single-SKU pages don't.

This is the whole game in D2C: the basket has to clear the cost to acquire it. The Discovery Set is the only major destination that does, a ~$113 set AOV against a $108 CPA, and it's the only one returning above 1.0x. Every single-SKU page sells an $80–90 basket it can't acquire profitably; Nightcap x Open is the extreme, an $80 order against a $145 CPA. The fix is the lever every scaled D2C brand eventually pulls: drive AOV with sets, bundles and subscription. Alice already has the scaffolding, several sets and bundles are live (Discovery Set, Bedroom Bundle, Big Night Bundle, The Essentials), the budget just isn't going there. Push spend and creative behind the sets, attach subscription, and basket size clears the CPA while the single-benefit drains get capped.

The plan closes two gaps at once. Tighten the creative to bring CAC down from $117, and lift the basket (Discovery Set mix, bundle attach, subscribe-and-save) so the breakeven floor moves up from $57 toward $65. CAC drops, floor rises, they meet in the middle. Breakeven ROAS holds at 1.67x in both scenarios (it's just 1 ÷ margin); only the CAC at which 1.67x is hit changes with the basket.

Subscriptions are the tell, and the upside. Of 6,882 orders in the last 90 days, 2,564 (37%) were subscriptions, worth $212K on the first charge. Here's the key: even before a single repeat charge, the first order should cover what it cost to win, and today it doesn't. Fixing that is the floor, and the plan stands on its own without counting a dollar of future value. The recurring revenue from those subscribers, which lives in Shopify and Recharge rather than Meta, is upside on top. Quantifying true payback by cohort is the first thing to pull next, and it can only make the case stronger.

Most of the setup is hand-built. The proven scaling path is automated.

The brief asked about account structure, so here it is in plain terms. Alice's campaigns are largely ABO (Ad Set Budget Optimization, where you set the budget for each ad set by hand and pick which ads run). The pattern across every DTC account we've seen scale past $300K to $3M a month is CBO (Campaign Budget Optimization, where Meta's algorithm allocates spend across the ads inside a campaign and lets the best ones win automatically). At Alice's spend level, the ABO setup is doing the algorithm's job by hand, which introduces human error and slows the learnings the platform is already capable of.

Numbers as Meta reports them, which run a touch higher (~0.92x blended) than the stricter 0.80x we use elsewhere, the attribution difference from earlier, not a discrepancy. *The Target row is in-store sales Meta can't see, not a true failure.

The signal is already there. The one Meta-automated campaign (Advantage+ "Old Legacy") carries the largest share of spend at 1.09x ROAS, while the hand-built campaigns sit in the 0.57 to 0.86 range. The single existing CBO campaign is under-funded and underperforming today, which means the move isn't "any CBO," it's a deliberate rebuild of the campaign structure around CBO with the right ad pool inside.

On retargeting. The account does retarget, but there's no deliberate incremental retargeting layer designed to bring engaged-but-not-yet-buying audiences back. Worth noting: even without a dedicated layer, Meta naturally delivers about 10 to 15% of any broad campaign's spend to people already engaged with the brand, so a "$0 retargeting" framing would be wrong. The opportunity is building a measured, incremental layer on top of that, not flipping a switch.

What we'd do. Consolidate spend toward Advantage+ (the proven performer at 1.09x), rebuild the rest of the structure under CBO so the algorithm allocates at the ad level, and design a small incremental retargeting layer once the base structure is right. The three audiences worth carving out are clear: cart abandoners, recent page viewers, and ad engagers (people who clicked or watched more than 50%), sized as a disciplined 10 to 15% of total spend. The order matters: structure first, retargeting layer second.

Nine cuts on the creative itself.

You can pull these nine cuts in Motion yourselves. The point isn't the data, it's that every cut lands on the same three moves: concentrate on the multi-benefit set shapes that already win, close the conversion gap with creative instead of discounts, and refresh a library that's gone thin and stale. The cuts below are the proof. Cut 1 is open; expand any of them.

The library, at a glance.

It looks varied, 32 of 37 styles are live, but the money isn't. Three styles hold roughly 70% of the spend, and the account over-produces its weakest video while its best styles sit nearly untested: Montage runs 38 ads at 0.69x and Greenscreen 20 at 0.52x, while Collage (1.16x) has 3 ads and How To (0.95x) and Pattern Interrupt (0.96x) just one each. That's a lot of labels, not a lot of real bets.

One ad can be tagged with more than one style, so the spend column adds up to more than the $809K total. Read each style's ROAS, not the column total. What's missing: 5 styles are completely untested (Founder, Skit, Meme, Transformation, Screen Recording) plus a barely-tried set (UGC Selfie, POV, Case Study). But the first move isn't the unknown, it's scaling the proven single-ad winners (How To, Pattern Interrupt) and the under-built statics (Collage, Offer-First). On angle and audience, lean into Relatability and Aspirational hooks (Cut 6) and the multi-benefit "wants it all" buyer (Cut 7), the shapes already converting.

A four-week rewire, read straight off the nine cuts.

Every move below traces to a cut. The format mix comes from Cut 2, the hook distribution from Cut 6, the kill list from Cut 5, the audience from Cut 7. No new spend to start, week one pays for the rest.

In plain terms. The Discovery Set banner and the "wants it all" buyer are the only things winning at scale (Cuts 1, 2, 7). The library is thin and getting tired (Cut 3), the hooks grab attention but don't close (Cut 4), and the worst per-dollar drain is using single-need SKUs (Nightcap, Brainstorm) to chase cold buyers when they convert better as retention drivers (Cut 5, plus the two-lane logic below). So the plan does three things: point the budget at the proven shape, redeploy the single-need SKUs into the retention lane where they earn, and rebuild video on the one format that actually converts. One more move sits outside the creative: restructure the account from ABO toward CBO so the algorithm allocates spend at the ad level, and design a measured incremental retargeting layer on top.

ACQUISITION

Party Trick · Happy Ending

Lifestyle, social occasion, intimacy. The eye-catching cold creative that pulls new buyers in.

BRIDGE

Discovery Set

The multi-benefit basket that already returns 1.04x at $113 AOV. Where buyers land and the basket clears CPA.

RETENTION

Nightcap · Brainstorm · Zen-X

Specific needs. Better as post-purchase + subscription drivers than as cold acquisition ads.

Two lanes, not one funnel. Today every SKU is fighting for the same cold-acquisition slot, and Nightcap-x-Open (the single-need sleep page) is the biggest drain at 0.55x. The honest read: Party Trick and Happy Ending are built for cold attention (occasion, social, intimacy), while Nightcap, Brainstorm and Zen-X convert better against a buyer who's already in. So the Cut 5 "kill list" isn't really a kill list for those SKUs, it's a redeployment list: the failing acquisition ads stay paused, but the SKUs move into the lane where they actually work, retention and subscription. That single shift unwinds most of the "single-benefit drain" story without losing the products.

What we'd tackle in the next pull.

Two tracks. First, the business context that makes the scaling decisions real. Second, the creative depth that pulls the account-moving levers.