Kalda jar sealer · plan written 2026-11-13 · data 10-21 to 11-12 · nothing launched, every Meta write waits on your go

Batch 5: what actually worked, and six bigger swings

We pulled the raw numbers (ad, text, headline, hour, age, placement), the Shopify orders, every comment on our posts, 22 scaling ads from TrendTrack checked against the live Ad Library, and 67 buyer quotes. Three of our old conclusions turned out to be wrong.

$38.73live CPA (10-24 to 11-12)

1.84Shopify MER vs 1.69 breakeven

72%of presell clicks never see the PDP

82%of purchases are women 65+

$61.6CPA 6 to 11am ET vs $31.7 rest of day

1. What the raw data says

Corrections to earlier verdicts

10-21 to 10-23 was a dead checkout. $588 spent, 0 Shopify orders. Clicks and add-to-carts were normal; only checkout broke (payments outage and the move to instabrush.co). The "N8 money hooks lost", "WTF images lost" and "loss/waste lost" calls were all made from that window, so they are not purchase evidence. WTF stays parked because of its PDP reach, not because of sales.
No creative difference is statistically real yet. Every ad with readable volume lands at a $35 to $43 CPA (N2 $38.8, D $34.6, N4 $43.2). D vs N2 in the same window: p = 0.68. "D beat control" is unread.
Meta gives 95% of each 3-hook ad to one hook. Sibling hooks got $35 to $39 in 20 days, so no hook comparison we ran was ever a real test. From now on: one hook per ad.

What is real

Concrete scenes pull buyers through the presell. Inside the same ad, the Walmart hook sends 29.7% of readers to the PDP and the mom's-garage hook sends 8.6% (p = 0.0002). Money-math hooks sit at 8 to 13%. The hook's measurable job is click quality.
The presell is the biggest leak. Only 28% of readers reach the product page. After that the funnel is healthy: 18% ATC, 68% to checkout, 62% complete. No ad has ever gone straight to the PDP.
Two media leaks: 6 to 11am ET is 39% of spend at ROAS 1.17 (the rest of the day runs at 2.26, and the gap shows up in both halves of the data). Women 55 to 64 run a $70 CPA.
Narrator mismatch. Every winning ad is told by a daughter in her 40s, but the buyer is a woman over 65. The buyer-age narrator (A, B, C) got $74 in total, so it is untested, not lost.
Economics: AOV $71.35, mix 39 single / 23 two-pack / 8 three-pack, breakeven ROAS 1.69 (CPA $42.25). The last 5 days ran at MER 1.76, roughly breakeven.

What the comments object to

Only 22 comments exist across 57 posts. The real objections: "how is an elder person supposed to unscrew the top?", "this is all reliant on having power", "so much cheaper on Amazon", "I bought a $25 one three years ago and my jars are still sealed". Reddit adds seal distrust: "it sealed only half the times and only if I put my whole weight on top of it".

2. Market: what is scaling that nobody in our lane runs

Mason Genie is still the only jar sealer scaling (100+ active ads, 31 launched in the last 7 days), and it sells "freshness" in video. Nobody runs long natives for 65+ in the jar lane. Searching transcripts for "weevils" or "bugs in my rice" returns 0 US video ads running 30+ days. The openings come from adjacent lanes:

Structure	Who / proof it scales	Kalda port
"Did everything right, missed one thing"	Blacksight, 57 days, 100+ ads on one landing page, 42 new this week	B5-3
Elder publicly bins the staple at the table	Neralle, 100+ ads all launched in the last 6 days	B5-4
"Lying seal": the expert checks your hands, not your recipe	Oh Yummy, 43 days, 83 ads from one page	B5-5
TV authority + every storage method fails	Hearthfold, 60 days, oldest ad 111 days	folded into B5-3
Plastic to glass, long native	Knighton, 317 days; Mason Genie video only	benched for wave 3

Blacksight stockpile shelf ad — Blacksight

7 of the 8 saved scaling images are plain phone photos of the failure, not product shots. That fits our calm-image rule.

3. The six swings

Each arm isolates one variable, runs one hook per ad and has a written falsifier. The images below were generated today (Nano Banana 2.1, Gemini-QA'd, calm scenes, no WTF slot).

B5-1funnelbiggest lever

Send the winner straight to the PDP

Same N2 img1 copy, image and headline. The only change is the link: product page instead of the pantry-story presell.

Why: 72% of the people we pay for drop on the presell. The 9 orders that started on the PDP had a $75.75 AOV. This has never been tested.
Isolates: Landing page.
Falsifier: Direct CPA ≥ presell CPA after both reach 5+ purchases.
Pair with: The F1 50/50 split on pantry-story (the deep plan's top on-site lever, still unlaunched).

Control N2 image — Existing control image (N2 img1), no new image needed

B5-2narrator

Same Walmart trigger, told by the buyer herself

The woman ahead of me at Walmart had four 20-pound bags of rice in her cart. I'm 68. I know what that means.

Headline: "I was the one stocking up in 1979"

Why: The hook that sells is a witnessed scene, but our narrator is 25 years younger than our buyer. Keep the trigger and swap the voice.
Isolates: Narrator age (V3). Body = D's body in her voice.
Falsifier: CPA > $42 or PDP reach < 25% at $210.

Cart of rice sacks — Cart ahead. QA flag: shows more than 4 sacks

Rice sacks and empty jars on kitchen table — Home after. QA flag: sack/jar count

B5-3new structureBlacksight port

"I stored it right and lost it anyway"

Two years of rice and flour on my garage shelf. I threw out 60 pounds in one afternoon.

Headline: "I never had a storage problem. I had a bag problem."

Why: The structure with the most volume in the scan. Buyers say it in our own words: "I've pitched about 500 lbs. of white flour over the decades", "the whole sealed bag in a sealed bin was completely infested".
Mechanism: The bag-to-bag spread, reframed: the prep was right, the container was wrong. One bad bag in one jar stays one bad jar.
Falsifier: PDP reach < 25% by day 3 (broken), or CPA > $42 at day 7.

Garage shelf of stored bags — The stockpile. QA 5/10: minor pegboard warping

Torn flour bag in garage trash — The afternoon. QA pass 7/10

B5-4witness + elder authorityThanksgiving: launch by 11-16

The aunt who dumped the rice at Thanksgiving

My aunt is 84. She opened my rice at Thanksgiving and poured it in the trash in front of everyone.

Headline: "Then she showed me her pantry"

Why: Neralle's version is the fastest-scaling ad in the scan (100+ ads in 6 days). It stacks all five stakes moves: witnesses, public, permanent, a cost you can see, and an elder authority who is the buyer's age.
Isolates: Public-shame witnessed trigger vs the stranger-at-Walmart trigger.
Falsifier: CPA > $42 at day 7. If it wins, holiday timing is a confound, so re-read in December.

Elderly woman pouring rice into trash — The moment. QA 5/10: one hand needs a fix before launch

Pantry of glass jars — Her pantry. QA pass 7/10

B5-5objection reversal

"Every lid was sealed. One wasn't."

Every jar in my pantry was sealed. I pressed the lids one night and one popped up.

Headline: "Press the middle of the lid. If it moves, read this."

Why: Seal distrust is the #1 objection to sealers ("sealed only half the times", "I wouldn't trust one", the $25-one comment). Make the objection the hook. The kit answers our own comments: opener in the box (for "how do I unscrew it"), USB-C charging, and a vacuum holds with nothing plugged in (for "reliant on power").
Isolates: Objection-first opener vs trigger-first.
Falsifier: CPA > $42 at day 7.

Domed jar lid close-up — The lid that moved. QA pass 7/10

Row of jars, one open — The one jar. QA 5/10: jar thread detail

B5-6new mechanism from buyer quoteswave 2

The freezer ritual

For 30 years I froze every bag of flour for four days before it went in my pantry. It didn't help.

Headline: "Freezing kills what's in the bag. Not what comes next."

Why: 83 buyer mentions of freezing every bag first ("I always freeze first now"), and zero ads run it. It names a chore she already does and gives her a reason it fails: the next bag from the store brings new eggs, and an open bag lets them spread.
Isolates: Failed-ritual mechanism vs trigger scene.
Falsifier: CPA > $42 at day 7.

Chest freezer full of flour and rice bags — The freezer. QA pass 7/10

4. Test design

Wave	Arms	Budget	Launch
Step 0	Graduate D img1 into the control as its own ad set, same post	control $120	on go
Wave 1	B5-1 direct PDP, B5-2 buyer narrator, B5-4 Thanksgiving aunt, B5-5 lying seal	ABO 4 × $30 = $120/day	by 11-16
Wave 2	B5-3 stored it right, B5-6 freezer ritual	ABO 2 × $30	after the wave-1 day-3 check

ABO, not CBO. Both CBO batches handed 75 to 91% of spend to one ad set. FB feed, broad Advantage+, Purchase on the Kalda pixel, utm_content={{ad.id}}.
One hook per ad, 2 images = 2 ads per arm. Archived alternates stay out.
Day 3 (broken-arm check): PDP reach under 20% with $60+ spent = broken, so fix or pull it.
Day 7 (verdict): win = CPA ≤ $42 with ≥ 5 purchases, confirmed against Shopify orders by day, not the pixel alone. Winners go into the control as same-post ad sets.
Media test, your call: a control duplicate scheduled to skip 6 to 11am ET. Ad scheduling needs a lifetime budget. The 55 to 64 CPA is noted, but no age exclusion, per your no-hard-coded-demographics rule.

5. How this maps to the 11-09 Batch 4 plan

Batch 4 item	Now
Step 0 graduate D	Kept. D's "win" is unread (p = 0.68), but it is tied for best and has had no spend since 11-07.
4A D-hooks (3 single-hook ads)	Replaced by B5-2 and B5-4. Both test the witnessed trigger, with bigger variables than swapping the Walmart detail.
4B / 4C reruns (Mom, Cashier)	Folded into B5-2: buyer-age narrator on the proven trigger, in its own budget.
4D N11 twelve-year jar	Deferred. It pushes V1 further, but V1 is unread.
F1 split	Kept, paired with B5-1.
Offer (3-pack only 8%)	Still open. Pick a gift on 2/3-packs or a post-purchase add-on once F1 has a read.

6. Before launch

Your go on Step 0, the wave-1 ABO at $120/day, and the F1 split.
Full bodies: write B5-2, B5-4 and B5-5 through the copy lane (doctrine, humanizer, QA) on D's body skeleton. The hooks above are drafts.
Edit pass on the 3 images flagged 5/10 (B5-4 hand, B5-5 jar thread, B5-3 pegboard).
Build paused, verify links, creative IDs and utm, then activate on go.

Sources: d1113/DECOMPOSITION.md, MARKET_SCAN.md, VOC.md and raw JSON in output/kalda-sniper-2026-10-31/d1113/. Not verified: Amazon and Walmart reviews (blocked), audience targeting of competitor ads, and session-level presell analytics (pixel proxies used).