Barton Springs Mill · Fluid IQ Digital

What's actually working in the creative, with the numbers to prove it.

147 ads, 18 months of spend, every top creative visually inspected against its own ROAS. A deterministic rule set for what to make next — not a gut call — plus what it means for Heartwell, the funnel, and where budget should move.

Prepared 19 August 2026 act_527467045489440 · Feb 2025 – Aug 2026 Read-only analysis — no account changes made

Coverage

147
distinct ads carried spend
in the 18-month window
$68,249
total spend analyzed
top 58 ads = 90% of it
81%
of spend classified
($55,001 of $68,249)
22
distinct creative concepts
visually inspected directly

The remaining 19% of spend is smaller one-off ads with no inspected sibling — excluded from the rule set rather than guessed at.

1 · The rules that hold — act on these

Largest, most confident split

Default new prospecting creative to no people in frame.

18.38x ROAS on 56 ads / $43.9k spend with no people, vs. 10.53x on 6 ads / $11.1k with hands only. Product-forward shots outperform process/human shots by roughly 1.75x ROAS at this sample size.

Weakest bucket — stop running these

Process shots with no finished product visible are the account's worst performer.

7.24x on 5 ads / $8.0k. If a person appears, they should be actively finishing a visible product — hands mid-process with no payoff is the one visual pattern to actively cut.

Workhorse angle

Technique / how-to copy is the account's largest classified volume at strong ROAS.

18.64x on 14 ads / $19.6k — the default angle for new copy unless deliberately testing something else.

Hero image choice

Finished baked goods beat raw ingredients as the lead image.

Baked good 19.17x (8 ads) vs. wheat berries 12.98x (6 ads) vs. flour bags 16.44x (6 ads). Bags and berries work as supporting/product-ID shots, but a finished loaf, cake, or cookie as the hero converts better.

AxisValueSpendROASn ads
peopleabsent$43,88818.38x56
peoplehands only$11,11410.53x6
subjectbaked good (finished)$11,52519.17x8
subjectraw bag (flour)$9,37616.44x6
subjectcatalog / DPA (feed) *$8,95022.31x29
subjectraw wheat berries$8,13512.98x6
subjectprocess (no finished product)$8,0487.24x5
subjectmixed (multiple in shot)$4,63016.06x3
subjectbag macro + meme text$4,33925.51x5
angletechnique / how-to$19,64818.64x14
angleprovenance / farmer-story$11,56214.07x9
anglecomparison (meme)$4,33925.51x5

* Catalog/DPA posts the second-best ROAS but pulls its image live from the product feed — no creative decision was made. This is a funnel-position effect (bottom-funnel retargeting on warm traffic), not a content signal, and is kept out of the "what to make next" rules below.

2 · Leads worth testing deliberately — not proven yet

Best number in the data, thinnest sample

"Dear grocery store flour, it's me not you" — comparison/meme copy on a bag macro shot.

25.51x on just 5 ads / $4.3k. This is the single strongest ROAS anywhere in the account, but on a sample this small it's a hypothesis, not a rule. Recommend 3-5 deliberate follow-up variations in the same format before giving it real budget share.

3 · Seasonality — real month-by-month data, not assumed

CategoryPattern
CakePeaks Sep–Oct: 11.95x (Sep '25) and 19.41x (Oct '25) vs. a flatter 4.8–10.3x Feb–May.
CornmealConcentrated Oct–Nov (Thanksgiving): $367→$1,197 spend, 4.63x→7.30x ROAS.
BreadYear-round workhorse with a summer bump — 19.75x in Aug '25, 13.7–15.3x Jun–Jul.
Pizza/flatbreadThin and inconsistent (0.5x–9.9x) — under-tested, not seasonal.
Holiday bundlesDegrades through December: 7.60x (Nov) → 3.55x (Dec), consistent with shipping-deadline pressure.

Actionable read: brief cake creative in August for a September launch, cornmeal in September for the Thanksgiving push, treat bread as the always-on baseline it already is, and don't scale pizza/flatbread on the current sample.

4 · What this means — Heartwell, funnel, and budget

01

Is Heartwell TOF?

Heartwell is a product line, not a funnel stage — it needs its own TOF-to-retarget path like any other SKU family. Its historical creative (process shots, provenance angle) sits in the account's weaker buckets even though the product converts fine downstream on retargeting. Launch Heartwell TOF with a finished-loaf hero shot and technique or comparison copy, running alongside the existing campaigns as its own line — not folded into one of them.

02

Funnel flow, top to bottom

TOF — finished-product hero, no people, technique or comparison angle, one campaign per active product line. MOF — DPA/retarget on view-content, add- to-cart and site-visitor pools (already the account's proven 22x+ bucket). BOF — existing-customer and cart-abandon retarget with tighter exclusion windows. This matches the account's own historical shape; the fix is what feeds TOF, not the funnel structure itself.

03

Budget adjustments

Leave the bottom-funnel retarget line as the account's biggest active spend — it's the proven bucket. Judge the Awareness/ThruPlay line on view-pool growth, not ROAS, since it isn't optimizing to purchase by design. Wheat Berry Mill and Flatbread haven't been re-tested against the strongest classified angle yet — worth a creative refresh test. Heartwell had zero paid spend as of this analysis; a modest test budget, sized to real fulfillment capacity rather than inventory count, is the recommended next move.