The Rebuild · the protein powder shelf

Levels whispers
its proof

Levels has the rarest thing on the protein shelf: claims that are true and verifiable. Grass-fed concentrate, real Madagascar vanilla, a third-party purity award. The main image prints all of it at whisper volume. Below is the entire job that fixed it, in the order it happened, including the three moves we refused to make.

the rebuilt main and gallery in an illustrative Amazon layout · rating as captured june 30, 2026 · chrome is a mockup, the creative is not

Brand

Levels Nutrition

ASIN

B071JG5QYT

Shelf

protein powder

Shelf capture

june 10, 2026

Read

67 winners · 532 images

Verdict

Real proof, printed at whisper volume

Ten stages

01The intake 02The brand kit 03The shelf, decomposed 04The read 05The listing, scored 06The decision log 07The render, and the gates 08The verdict, at 240px 09The gallery 10What it does not do

Stage 01

The intake: what the product is allowed to say

Before anyone looks at a shelf, the product gets a brief. Every claim, certification, ingredient and printed line goes into one list, and that list is a hard boundary. Nothing enters the rebuilt image that is not in it.

The brief · Levels Grass Fed Whey Protein, Vanilla Bean, 2LB

Claims

Each scoop contains 24g protein, 5.4g BCAAs, and no added sugars.
NOTHING FAKE: no hormones, no GMOs, no added sugars, no soy, no gluten, no bleach, no fillers, no artificial flavors or sweeteners.
Made with Madagascar vanilla extract, lightly sweetened with monk fruit extract.
Grass-fed hormone-free dairy only; whey concentrate is the least processed form of whey.
Made in the USA, undenatured, cold-process microfiltered, blended with non-GMO sunflower lecithin.

Ingredients, all six of them

Whey protein concentrate · Vanilla extract · Sunflower lecithin · Sea salt · Stevia leaf extract · Monk fruit extract

read off the panel the brand prints in its own gallery, and corroborated by the pack seal reading ONLY 6 INGREDIENTS

Printed on the package

24G of Protein5.4g BCAAsNo Added SugarsNo ArtificialsGrass FedNo HormonesNo GMOsNo SoyNo GlutenNo BleachNo FillersMadagascar Vanilla ExtractMonk Fruit ExtractMade in the USAUndenaturedCold-Process Microfiltered

claims and printed lines transcribed from the live listing and the product's own pack

The governing rule

Never claim what the product does not hold.

Every number, claim and certification in the rebuilt image has to trace back to the list on the left. Not to the shelf, not to a competitor, not to what would sell better. Stage 06 has a decision where this rule cost the image a good idea.

Facts panel, as supplied

Protein24g

BCAAs5.4g

Added sugars0g

Servings per container28

Stage 02

The brand kit: what it is allowed to look like

Stage 01 bounds what the rebuild may say. This is the same governor on the other axis. The kit is read off the brand's own pack and site, and it earns its keep everywhere there is no original to copy: the one added device, and every slide of a gallery that does not exist yet.

Palette, taken from the pack

Clean White#FFFFFF · primary

Ink Black#1A1A1A · ink / text

Vanilla Butter#EBD98A · accent

Golden Mustard#E4B23E · accent

Strawberry Coral#D96A66 · accent

Chocolate Brown#9A6233 · accent

Deep Forest Green#2E3B2E · neutral

Off-White Paper#F2F0EB · neutral

The rest of the system

Headline type

Heavy condensed sans, all caps, tightly stacked, squared terminals. Industrial athletic.

Numerals

Bold weight stat callouts (24G, 5.4G) contrasted against lighter labels.

Motifs

Circular seal callouts, grayscale grass-fed cow, bottom flavor banner with a diagonal cut, laurel award emblem.

Photography

Real everyday people, sunny parks and lived-in kitchens, warm retro film grade with golden highlights.

Voice

Confident, clean, slightly cheeky. Minimal ingredients, purity, no nonsense.

the receipts are two stages away: the roundel is charcoal and cream rather than a stock gold medallion, and the rebuilt gallery runs their cow, their laurel and their flavor band rather than the generic supplement look

On the main image this matters less than it looks, because the pack itself is frozen and the label is reproduced pixel for pixel. A shopper who has bought this tub before will recognize the one in the rebuilt image, and stage 07 is where that gets proved rather than promised.

It matters enormously everywhere else. The added roundel and all six gallery slides are new artwork with no original to be faithful to, and that is precisely where a rebuild turns into a rebrand nobody asked for. Left to its own taste, a generator reaches for the category default: gold foil medallions, stock athletes, a palette the brand never chose.

The rule, stated as a rule

We are hired to merchandise a brand, not to redesign one.

Nobody has measured what a shopper feels when the listing and the box that arrives look like two different companies, and we are not going to pretend otherwise. It is an operating rule here, not a finding: the kit is the boundary, and every asset in this case file was held inside it.

Stage 03

The shelf, taken apart image by image

67 winning listings on the protein powder shelf, captured June 10, 2026. Every main and every gallery slide decoded into a fixed schema, element by element. This is counting, not critique, and the separations below are the part that makes it counting.

winning listings

532

images decoded

465

gallery slides

8,354

elements scored

Five things that get confused for each other

on_package

Text PRINTED on the real label. Typed by role, sized, placed, and flagged if it is the pack's dominant element.

composited_overlays

Graphics ADDED on top of the photo. Badges, stat discs, callouts. Never confused with printed copy.

on_package_imagery

Illustrations and textures printed on the pack. The cow is artwork, not a prop.

staging_props

Real physical objects placed in the shot. Scoops, pods, a poured glass.

facts_panel

Supplement panel rows, read line by line.

a badge stuck on the photo and a claim printed on the label are different facts, and a shelf read that mixes them is worthless

What one element looks like when it is scored

{
  "text": "WHEY PROTEIN",
  "element_type": "product_name",
  "printed_vs_composited": "printed",
  "region": "center of label",
  "placement_pct": { "x": 68, "y": 52, "w": 18, "h": 7 },
  "relative_size": "hero",
  "typography": "bold condensed sans, uppercase, two lines",
  "is_hero": true
}

one of the elements scored on this listing's own main image · the same schema runs on all 67 winners, which is what makes the comparison in stage 05 an apples-to-apples one

Stage 04

The read: what the shelf has settled, and what it has left alone

One question per element: how many winners do it? High adoption is an assignment, because the shelf already ran that test. Low adoption is an opening. Absence is never a prohibition.

The read · protein powder · 67 winning main images

element	winners	share	pile
Legible at thumbnail size	67 / 67	100%	convention
Something printed at hero scale	67 / 67	100%	convention
A numeric stat printed on the pack	67 / 67	100%	convention
Photographed, not rendered	61 / 67	91%	convention
Clean packshot, no staged scene	49 / 67	73%	convention
Product fills 70% of the frame	median	70%	convention
Certification or trust mark in frame	39 / 67	58%	convention
A flavor or ingredient prop staged	9 / 67	13%	white space
A person in frame	3 / 67	4%	white space
A prepared drink in frame	1 / 67	1%	white space

counted image by image off the 67 winning mains · an element on every winner and an element on three of them are not close calls at this sample size, which is why the table sorts rather than ranks

The finding that runs this rebuild

The product name39 / 6758%

The brand wordmark26 / 6739%

A benefit claim1 / 671%

An ingredient callout1 / 671%

A number0 / 670%

the pack's single dominant element, counted across all 67 winners · 42% do print a stat at hero or large size, but never as the biggest thing on the pack

Every winner on this shelf prints a number somewhere. Not one of them makes a number the thing you see first. The biggest element on a winning protein pack is the product name or the brand, and only 42% print their stat at hero or large size at all.

That is the opposite of what the category is assumed to do, and it means the rebuild's central move, heroing 24G, is not a convention match. It is a deliberate departure. Stage 06 is where it gets argued rather than assumed.

A read is only worth running if it is allowed to tell you that your instinct is the minority position.

Stage 05

The listing, scored by the same schema

Only now does the subject get looked at, and it gets the identical treatment: every element typed, sized, placed and graded, then diffed against the shelf. Taste never enters it.

The current main · B071JG5QYT

Levels Grass Fed Whey current Amazon main image: white tub with cow artwork, 24G protein in small gold text, four small seals along the lower label edge

strongweakabsent elements carry no pin

Element scoring · nine slots, every listing

1Brand / wordmarkstrong

LEVELS wordmark clear and dark on the label top left.

2Ingredient / source cuestrong

Cow graphic signals grass-fed dairy whey source.

3Hero stat or numberstrong

24G PROTEIN and 5.4G BCAAS in gold, legible, but sized medium against a hero-sized product name.

4Primary benefit claimweak

'Minimal ingredient protein that helps build lean muscle' is small low-contrast gray text.

5Certification / trust markweak

Four seals present, small, and the last one is cut by the curve of the jar.

6Clinical / proof cueweak

The Clean Label Purity Award seal implies third-party proof but is tiny.

Lifestyle / in-use cueweak

Cow imagery suggests origin, but nothing shows the product in use.

Texture / formula cueabsent

No powder, no scoop, nothing that shows what is in the tub.

Social proofabsent

No ratings, reviews or endorsements.

fill 55% against a 70% shelf median · legible at thumbnail · Amazon compliant

The label, transcribed before anything is changed

Wordmark

LEVELS followed by a registered mark, a circled R. Confirmed by magnifying the glyph 30x: an enclosed ring, not the wide flat pair a TM would produce.

Footnote key

'THAT HELPS BUILD LEAN MUSCLE' ends in a double dagger, two crossbars, not a single dagger. Read off a raw pixel dump.

Seal 1

ZERO ARTIFICIALS

Seal 2

ONLY 6 INGREDIENTS

Seal 3

clean label PURITY AWARD, inside a laurel wreath

Seal 4

28 SE… , a servings claim, cut off by the curve of the jar

the rebuild treats the label as fixed pixels, so it has to know exactly what is on it first

Not readable, and therefore not guessed

The flavor-band sub-line beneath VANILLA BEAN. White type on pale yellow, visible as a shape at 16x, no glyph separable.
The word PROJECT inside the purity seal, and the mark that follows it: registered, trademark or none.
The second line of the servings seal. Only 'SE' sits on the visible face; the rest wraps around the jar.
The net weight figure, and the characters after (32 OZ). Both off the visible face.
Both fine-print legal footer lines over the cow photograph.

these regions are reproduced as pixels and never re-typeset · a rebuild that invents the fine print it could not read has invented a claim

The gaps, with severity

highProduct is small for the shelf

55% fill against a 70% shelf median

A bigger pack reads faster at thumbnail and signals confidence.

medProof cue is weak

A third-party purity award, printed at seal size

58% of winners carry a trust mark. This brand has one it actually earned and prints it smaller than its own net weight.

medNo formula cue at all

Absent

Nothing in frame shows powder, scoop, or what a prepared serving looks like.

The listing is not broken. It is quiet. Its stats are printed at medium while its product name is printed at hero, its four seals are the size of a fingernail, and the fourth one slides off the curve of the jar.

Stage 06

The decision log

Nine decisions, each one recorded the same way: the shelf fact, where this product stands against it, what was chosen, and why. Three of them are refusals, which is the part a portfolio never shows you.

Hero the 24G the pack already prints, enlarged on the label itself

The shelf fact

Not one of the 67 winners makes a number the pack's dominant element, so a number hero is a departure here, not a convention. Protein is still the most-printed stat type, 22 of 67, and the claim route is overwhelmingly on-package: 95% of winners print, 3% badge.

Where Levels stands

The pack prints 24G PROTEIN in the brand's own gold and charcoal, at roughly a tenth of the label height. Sized medium. It is mush at a 240px search thumbnail while the shelf leaders print their grams three to four times that scale.

What we chose

Take the brand's own existing figure and re-set it dramatically larger in the label's own condensed type and colors, flat into existing label whitespace, following the curve of the tub.

Why

This is a deliberate move against the shelf's own habit, and the reason is in the product, not the pattern. The shelf is identity-led, so the read says do not force a number hero. But 24g from grass-fed concentrate sits at or above the norm here, the isolate leaders print 24 to 25 and the plant leaders 20 to 21, so this is a number the product genuinely owns rather than a table-stakes dose. Printing it on-pack rather than badging it keeps the highest-trust medium and the shelf's 95% route.

Exactly one added device: a charcoal ZERO ARTIFICIALS roundel

The shelf fact

Trust marks appear on 58% of winning mains. Composited badges are rare here, 3%, so a badge is a differentiator when justified. Coverage warning: 5 of the 6 composited badges on this shelf sit over the product.

Where Levels stands

The brand's whole wedge is NOTHING FAKE, on a shelf whose volume leaders print 'artificially flavored' on the front. The pack does carry a ZERO ARTIFICIALS seal, one of four small roundels crushed against the bottom label edge beside the cow. Invisible at thumbnail.

What we chose

Promote that one claim into a single prominent roundel in the brand's own deep charcoal with cream lettering, placed in the tub's clean white shoulder at upper right, overlapping the silhouette but off the label copy.

Why

A priority word claim that only fits on-pack as micro-type needs a badge, not a smaller font. Charcoal on cream is the product's own palette and deliberately distinct from the gold seals already printed, so it reads as an addition rather than a duplicate. Siting it in the white shoulder rather than across the label answers the coverage risk the shelf shows.

No added badges for Grass Fed, Non-GMO or Made in USA

The shelf fact

The trust lever exists on this shelf. That does not make every trust claim a priority for this product.

Where Levels stands

GRASS FED is already the second-largest line of type on the front label, and the cow photograph carries the same message wordlessly. Non-GMO and Made in USA are printed on-pack too.

What we chose

Carry all three natively. Zero added badges for them.

Why

Re-badging a claim the pack already prints prominently is clutter, and it would break the single-device cap. Ruthless hierarchy beats maximal inclusion.

Stage the flavor world: vanilla pods, one orchid bloom, one full glass of the prepared shake

The shelf fact

A flavor prop appears on 13% of winning mains but 78% of winning galleries. A prepared drink appears on 1 of 67 mains and 52 of 67 galleries. Both are openings on the main surface, not conventions.

Where Levels stands

The flavor is Vanilla Bean, made with real Madagascar vanilla extract, so the cue is unmistakable and truthful. The current main is a bare jar with no appetite cue at all.

What we chose

Three glossy pods, one split to show the seed paste, a single cream-white orchid bloom, and one full tumbler of blended shake set just behind them. Two prop groups, one side, the whole right of the frame left clean.

Why

The pods and the bloom are the instant vanilla signal and the glass sells the made product rather than the powder. The shake is specified in its real color, creamy off-white, never tinted vanilla yellow. Props stay grouped on one side because scattering them across the frame reads cheaper and destabilizes the render.

Liquid splash: dropped, logged as an A/B candidate

deferred to the test pile

The shelf fact

A liquid splash appears on 5% of winning mains, above the floor at which we treat a lever as validated, and it is truthful for a dairy shake.

Where Levels stands

The scene already carries a prepared shake and a prop cluster on the same side of the frame.

What we chose

Drop it from the image. Log it as a test candidate.

Why

A splash plus loose bits on top of the props is the exact move that crowds this frame and pushes elements toward the edge. It loses the conflict against a scene whose dynamic element is already the filled glass.

Servings count: refused until it could be sourced

deferred to the test pile

The shelf fact

Servings is the second most common stat type on this shelf, 17 of 67 winners. A supply stat on a 2LB tub would have been an attractive supporting lever.

Where Levels stands

At the time of the run the figure was not in the verified brief, and on the main image the servings seal wraps around the curve of the jar, so it could not be read off the product either.

What we chose

No servings figure anywhere in the image. Logged as a candidate for once the count is confirmed.

Why

The claims governor is absolute. An unverified number does not go on a client's pack, however likely it is to be right.

follow-upSince confirmed at 28 per container, sourced from the pack's own back panel, which the brand photographs at gallery slide 7. It is now a verified lever available to the next run, and it got there by finding a source rather than by assuming one.

No variant or flavor-range row

deferred to the test pile

The shelf fact

A variant signal appears on 5% of winning mains.

Where Levels stands

The listing has 23 variations across six flavor families and four sizes. A genuinely deep range.

What we chose

Dropped to the test pile.

Why

The range row is gated off by a human toggle that was set to off for this run. The flavor source props are not covered by that gate and stay in.

Hold the clean format, and freeze the product

The shelf fact

73% of winning mains are clean and Amazon-compliant. A person appears on 4%. Package density averages 15 printed text elements per pack on this shelf.

Where Levels stands

The tub is a single upright unit shot straight on, like every other main on the shelf.

What we chose

Pure white sweep, soft key from the upper left, real contact shadow, same unit in the same pose, angle and count. The scene description covers background, lighting and props only.

Why

The product is frozen: only the context around it changes. The scene deliberately does not re-transcribe the label, because enumerating label elements makes the renderer repaint the label from words and silently drop the fine detail, the seals and the cow.

A legibility budget: two added claims, both large, nothing else

The shelf fact

Two or three added claims is the ceiling that stays readable at 240px, and this shelf already averages 15 printed text elements per pack.

Where Levels stands

This label carries a headline, a stat pair, a sub-line, four seals, a flavor band and a net-weight block before we add anything. Its own copy is already competing with itself.

What we chose

One dominant on-pack stat plus one roundel. Nothing else added.

Why

The existing sub-copy eats most of the budget, so everything else goes to the gallery rather than being shrunk onto the main: 5.4g BCAAs, no added sugars, monk fruit, undenatured, cold-process microfiltered. That routing decision is what stage 09 is built from.

Decision 06 is the one worth reading twice. A servings count would have been a free lever on a 2LB tub, the shelf's second most common stat, and it stayed off the image until someone could point at where the number came from.

Stage 07

The render, and the four gates it has to clear

The decisions become a written spec, the spec becomes an image, and the image gets reviewed against the real product before anyone sees it. Failing any one gate sends it back regardless of how good it looks.

From the spec the image was built from

LABEL FIDELITY (HIGHEST PRIORITY): treat the product's
printed label in the attached photo as FIXED PIXELS to be
reproduced exactly, not text to be re-typeset. Keep EVERY
existing label element in its exact position, size and
proportion, including any emblem, medallion, seal,
certification icon, divider rule and fine print.

REGIONS THAT WERE NOT LEGIBLE in the reference: reproduce
those pixels as they appear and invent no text for them.

CLAIMS GOVERNOR: render only text, numbers and claims that
are printed on the real package or listed in the overlays
above. Invent nothing.

three excerpts from the run's spec · the full document runs to roughly 2,500 words and covers geometry, lighting, containment, the two permitted claim treatments and a final self-check

The four gates

Containment

If any element, prop, shadow, reflection or fragment touches or crosses any edge, the image fails. No exception, however good it looks.

Plausibility

Every claim must sit on a real package surface: printed on the label, on a badge attached to the pack, or on a hang-tag. A free-standing card propped beside the product is an instant fail. So is floating text.

Legibility

The reviewer looks at the 240px thumbnail. If a claim the plan called a lever is mushy at that size, it did not land. There is no borderline pass.

Fidelity

The render is compared against the real product photo band by band. Any dropped seal, re-lettered mark, reshaped container or swapped symbol fails.

a gate failure is a hard fail, not a deduction · the reviewer sees the render, the real product photo, and the 240px thumbnail

Three attempts, one promoted

Attempt 1

The roundel prints NO ARTIFICIAL SWEETENERS. The spec permits one string, ZERO ARTIFICIALS, and a reworded claim fails even when it is true.

Attempt 2

Wording correct. But 24G and PROTEIN break onto separate lines, and the roundel is sized to compete with the hero rather than support it.

Attempt 3 · carried forward

24G PROTEIN reads as one line at hero scale, the roundel sits clearly secondary, and the props stay grouped on one side of the frame.

The fidelity check

Fidelity check: a horizontal band of the real product label above the same band of the rebuilt render, aligned for comparison

real label on top, render below, same crop · run as three bands down the pack so a dropped seal or a re-lettered mark cannot hide

The image is generated, and the interesting question is not whether that is allowed. It is what stops a generator from quietly improving the label: turning a registered mark into a TM, dropping a seal it thought was clutter, inventing the fine print it could not read.

The answer is that the label is treated as pixels rather than text, the unreadable regions are named in the spec as off-limits, and the result is compared band by band against the real photograph. That is not theory. The first attempt above rewrote a claim into words the product had not been given, which is precisely why it stayed an attempt.

Anyone can make a prettier picture of a protein tub. The work is making one the brand could legally print.

Stage 08

The verdict, at the size it gets made

Both images have been on this page at full size, which is the one size no shopper will ever see them at. Here is the rebuilt main back in the row it has to win, against the real page one it was read from.

At this size the current tub is a white jar with a cow on it. The rebuild reads as 24 grams, grass fed, vanilla, and nothing fake. The proof finally carries.

protein powder⌕

their main image:

Amazon's Choice

Optimum Nutrition Gold Standard 100% Whey Protein Powder, …

★★★★★99,070

$4400

Optimum Nutrition Gold Standard 100% Whey Protein Powder, …

★★★★★99,070

$9628

Optimum Nutrition Gold Standard 100% Whey Protein Powder, …

★★★★★29,506

$7999

Premier Protein Powder, Chocolate Milkshake, 30g Protein, …

★★★★★28,176

$2597

Orgain Organic Vegan Protein Powder, Vanilla Bean - 21g Pl…

★★★★★61,779

$2959

Transparent Labs Grass-Fed Whey Protein Isolate - Naturall…

★★★★★1,655

$5999

rebuilt

Levels Grass Fed Whey Protein Powder, No Artificials, 24G …

★★★★★6,805

$8999

Muscle Milk Genuine Protein Powder, Chocolate, 1.93 Pounds…

★★★★★5,872

$2318

the captured row · "protein powder"

← protein powder

the main

the stat hook

the proof

the crave

the process

the routine

the close

1 / 7

Levels Grass Fed Whey Protein Powder, No Artificials, 24G of Protein, Vanilla Bean

★★★★★6,805 ratings

$8999

Add to Cart

every slide above is the rebuilt sequence, labeled by its job

the detail page · swipe or tap the rebuilt gallery

page one as captured june 30, 2026 · positions, titles, prices and ratings as observed · sponsored rows omitted · levels appears at its captured position; the toggle swaps only its main image · detail-page mock is illustrative amazon chrome, not amazon's page

Nothing was added that the pack does not already own. The same four seals, the same cow, the same flavor band, the same 24G figure. All that changed is which of them the shopper can see before they decide.

Stage 09

The gallery, read the same way

The main image wins the click. The gallery wins the sale, and it gets the identical treatment: the shelf's own composition counted first, then this listing's eight slides scored against it, then a rebuilt sequence where every slide has one job.

What a winning gallery on this shelf contains

Lifestyle43% · 199 slides · levels 3 of 8

Infographic33% · 155 slides · levels 3 of 8

Facts panel17% · 79 slides · levels 1 of 8

Comparison4% · 19 slides · levels none

Product on white2% · 11 slides · levels 1 of 8

465 slides across 67 winning galleries · median gallery runs 7 slides, range 4 to 11

This is where the read refuses to give an easy answer. Levels runs eight slides against a shelf median of seven, and its mix is close to the shelf's: three lifestyle, three infographic, one facts panel, one packshot. By composition, the gallery is doing its job.

The defect is placement. Three of the eight fail the same thumbnail legibility test the main image is held to, and the brand's hardest evidence sits on two of the three: the six-item ingredient panel appears at slide 2 and again at slide 7, and neither one reads. The proof is in the gallery twice, and unreadable both times.

Current · 8 slides, scored

1 · infographic

NOTHING FAKE, EVER

legible

2 · infographic

the ingredient panel and nutrition facts

not legible

3 · lifestyle

REAL TASTE

legible

4 · lifestyle

24G OF PROTEIN per scoop

legible

5 · infographic

5.4G OF BCAAS per scoop

legible

6 · lifestyle

ONLY SIX INGREDIENTS

legible

7 · facts panel

the pack back: ingredients, nutrition, amino acids

not legible

8 · product on white

the pack back: the NOTHING FAKE story

not legible

purpose and thumbnail legibility as scored in the shelf corpus, by the same pass that scored the 67 winners

Rebuilt · six slides, six jobs

Every slide is planned before it is made: an archetype from the shelf's own mix, the one value proposition it carries, the exact text it may print, and a review afterwards against the same rules the main image answers to.

1 · the stat hook

archetypesplit stat + lifestyle

the one thing it carries

24g protein and 5.4g BCAAs, no added sugar, sweetened only with monk fruit

text it is permitted to print

24G PROTEIN / Per Scoop

review · passAll required text present and verbatim. No invented badges or off-brief props. Reads at thumbnail.

2 · the proof

archetypecomparison

the one thing it carries

Six ingredients, named and shown, against a competitor label crammed with additives

text it is permitted to print

ONLY 6 INGREDIENTS. THAT'S IT. / the six panel rows, verbatim

review · passEvery row matches the printed ingredient panel in panel order, and the headline uses the pack seal's own wording. No product in frame, no badges, nothing invented.

3 · the crave

archetypeflavor cue

the one thing it carries

Made with real Madagascar vanilla extract, not vague natural flavor

text it is permitted to print

VANILLA BEAN

review · passLabel copy on the jar is legible package text, not a misplaced field label. On-brief for a flavor cue.

4 · the process

archetypespec callout

the one thing it carries

Cold-process microfiltered, undenatured whey concentrate, the least processed form

text it is permitted to print

COLD-PROCESSED WHEY / NEVER OVERHEATED

review · passAll required text verbatim and correctly spelled. Diagram uses the brand palette. No invented props.

5 · the routine

archetypelifestyle moment

the one thing it carries

Crafted for everyone's lifestyle, athletes and daily grinders alike

text it is permitted to print

FITS YOUR DAILY GRIND

review · passBranded tub and shaker read clearly. No invented badges or stat coins.

6 · the close

archetypebenefit icons

the one thing it carries

No hormones, no GMOs, no soy, no gluten, no fillers, no artificial flavors or sweeteners

text it is permitted to print

NOTHING FAKE. EVER.

review · passThe GRASS FED cow seal is brand imagery, not a fabricated award. Headline reads at thumbnail.

Stage 10

What the rebuild does not do

The constraints are the product. Anything that loosens them makes a better picture and a worse deliverable.

It does not invent a claim. 24G and 5.4G BCAAs are the label's own numbers, the four seals exist on the current pack, the vanilla is the named flavor, and every ingredient named is read off the panel the brand prints.
It does not repaint the label. Every emblem, seal, footnote key and column of fine print is reproduced from the reference, and the regions that could not be read are reproduced as pixels rather than re-typeset.
It does not add a number the brief does not carry. The servings count was refused for exactly this reason.
It does not claim measured lift. Every percentage on this page describes what winners on this shelf run, not what caused them to win.

Want this on your catalog?

A full shelf audit runs this same read on your listing and hands you a prioritized fix stack, not a checklist.

Request an audit →one shelf · one report · turnaround in 5 days

← All rebuilds The method behind this →Read the research →

decoded from 67 winning listings on the protein powder shelf, 67 main images and 465 gallery slides, captured june 10, 2026 · every share on this page is a count from that corpus, recomputed image by image · shares describe what winners run, not what causes winning · no A/B result is claimed anywhere on this page · the rebuilt main and gallery are generated images, reviewed against the real product photograph before promotion · ASIN B071JG5QYT

Levels whispersits proof

The intake: what the product is allowed to say

The brand kit: what it is allowed to look like

The shelf, taken apart image by image

The read: what the shelf has settled, and what it has left alone

The listing, scored by the same schema

The decision log

Hero the 24G the pack already prints, enlarged on the label itself

Exactly one added device: a charcoal ZERO ARTIFICIALS roundel

No added badges for Grass Fed, Non-GMO or Made in USA

Stage the flavor world: vanilla pods, one orchid bloom, one full glass of the prepared shake

Liquid splash: dropped, logged as an A/B candidate

Servings count: refused until it could be sourced

No variant or flavor-range row

Hold the clean format, and freeze the product

A legibility budget: two added claims, both large, nothing else

The render, and the four gates it has to clear

The verdict, at the size it gets made

The gallery, read the same way

What the rebuild does not do

Levels whispers
its proof