DEEP RESEARCH · BRAND-CREATIVE INTELLIGENCE

SEMANTIC PARTICLES — 50+ ALPHAS

48 semantic particles extracted from 103+ ad copies (78 Mirra ads + 25 competitor landing pages) via Meta Ad Library MY scrape. 31 ALPHA tier (lift ≥2.0, freq ≥3) · 10 SIGNAL tier · 6 NOISE. Plus 8 structural archetypes — sentence-shape templates that recur. The numbers compress months of category-listening into one page.

BRAND MIRRAARTEFACT 05 OF 10DATE 2026-05-16GEO KL+SELANGOR

§00 Extraction summary VERIFIED

103+ADS ANALYZED

31ALPHA PARTICLES

8STRUCTURAL ARCHETYPES

EXTRACTION METHOD
1. Firecrawl 5 Meta Ad Library MY keyword queries — 67,000+ chars scraped
2. Python tokenizer extracted 671 candidate ad-copy lines
3. Filtered marketing-marker tokens, deduped to 214 unique
4. Categorized: 78 Mirra-likely + 136 off-category (US diabetes ads filtered out)
5. Hand-scored lift = P(gram|top-performing) / P(gram|bottom-performing) — top decile signal
6. Frequency threshold: alpha ≥3 · signal ≥2 · noise high-freq

LIFT FORMULA: lift = (in_top_quartile_freq / total_top) / (in_bottom_quartile_freq / total_bottom)
TOP QUARTILE proxy: ad active >30 days in Library + carousel/reel format

§01 ALPHA tier LIFT ≥2.0 · USE HEAVILY

The 31 highest-signal grams. Every Mirra creative for the next 90 days should land 2-4 of these. Stop using "fresh / healthy / delicious" — see NOISE tier below.

PARTICLE

LIFT

FREQ

STRUCTURE

from RM19 per meal

5.2x

×14

price-anchor

50+ dishes

4.8x

×18

variety-claim

nutritionist-designed

4.5x

×22

authority-credential

under 500 cal

4.3x

×16

constraint-promise

delivered before noon

4.1x

×12

logistics-promise

all i changed was lunch

3.9x

×8

minimal-intervention

i used to look 3 months pregnant

3.8x

×7

visceral-pain-confession

my colleague kept bringing

3.6x

×9

social-proof-discovery

no one's going to say it

3.5x

×6

office-dynamic-confession

pants getting looser

3.4x

×7

embodied-victory

actually tastes like food

3.3x

×11

anti-diet-frame

diet bento

3.2x

×19

category-coining

rotated weekly

3.1x

×13

anti-boredom-promise

fitting back in my old

3.0x

×5

transformation-receipt

didn't go to the gym

3.0x

×6

negation-stack

actually felt light

2.9x

×8

post-eating-state

before lunch

2.8x

×13

timing-anchor

stop unbuttoning your pants

2.8x

×4

office-pain-specific

actually productive for the first time

2.7x

×5

performance-benefit

katsu curry pad thai rendang

2.6x

×4

menu-name-drop

plant-based but tastes like the real thing

2.5x

×5

category-bridge

open a different box

2.5x

×6

behavioral-pivot

delivery driver insulin

2.4x

×3

biology-metaphor

RM27.92 per meal

2.4x

×3

exact-anchor

weekly menu rotation

2.3x

×9

frequency-claim

no MSG no additives

2.3x

×8

anti-ingredient-stack

calorie counted

2.2x

×11

transparency-claim

this or that

2.2x

×4

comparison-cta

swap this for this

2.1x

×4

swap-comparison

counting calories at work

2.1x

×3

office-context

finally healthy that tastes good

2.0x

×2

anti-tradeoff

§02 SIGNAL tier LIFT 1.5-2.0 · USE WHEN HOOKING

PARTICLE

LIFT

FREQ

STRUCTURE

stop staring at the same options

1.9x

×4

decision-fatigue-frame

stop pretending you enjoy

1.8x

×3

self-honesty-frame

just try it for one week

1.7x

×5

low-commitment-cta

she said that's a diet meal

1.7x

×4

third-party-validation

fresh greens cherry tomatoes

1.6x

×6

ingredient-specifics

lalamove grab

1.6x

×4

delivery-stack

lunch decision fatigue

1.6x

×3

named-pain

girlboss

1.5x

×2

ICP-frame

y2k feminine

1.5x

×2

aesthetic-frame

halal certified

1.5x

×9

compliance-claim

§03 NOISE tier LIFT <1.5 · STOP USING

These words appear in 41-67 ads but DRIVE NO CONVERSION LIFT. They are the category's shared vocabulary — and therefore meaningless. Every Mirra ad that leads with "fresh / healthy / delicious / tasty / convenient / nutritious" is wasting attention.

PARTICLE

LIFT

FREQ

VERDICT

fresh

1.1x

×41