An eye for clothes. It answers your team. It dresses your shopper around the piece they're looking at.
You give it your catalog. It gives your shopper a complete look, and the shelf of pieces around it, built on what they're looking at right now — and it keeps no profile of them. No cold start. No tracking script on your site.
Most tools give your team a dashboard and your shopper a widget, and the two never meet. Bunsar is one eye, on both surfaces.
Your team asks it questions in plain language and reads plain-language answers back. Your shopper gets a complete outfit, built on demand, from the same eye. There is no export, no nightly copy, no second model that drifts out of sync. Both sides read the same live catalog, the real thing.
And when the range itself is thin, it says so — in that same plain language, and it names the pieces. A product it cannot confidently build a look around. An outfit it cannot finish because a garment has no photograph. You will watch it do exactly that a few screens down. It surfaces the problem and hands it to a person. It does not quietly change your setup.
Striped Wool Crewneck SweaterFrom our own demo catalog
The surface
This is what your product page looks like with Bunsar on it.
Not a diagram of the output — the surface itself, rebuilt. This is a reconstruction, not a screenshot
of someone's shop: a retailer's product page built here, on our own demo catalog, with the Bunsar
bands in the places they actually occupy. The grey part is the shop's. Everything under the line is
ours, and each band carries a credit naming what it is, which recommendation its contents came from,
and whether it served them or we composed them — four of the five ship exactly like this;
the fifth, the feed, is ours to compose, and its credit says so.
store.example/product/BNS-C653A251412E
Atlas & Co
Striped Wool Crewneck Sweater
BNS-C653A251412E · top
$84.00
XSSMLXL
Add to bag
The retailer's page. Everything below the line is Bunsar.
everything below is served by Bunsar — 5 widgets, sliders and feeds the shopper pushes, nothing computed while they wait
Style suggestions
bunsarcuratedOutfitscomplete outfits, not lookalikesserved, not computed on the click
8 ways to sell this one product as an outfit, in the order the engine ranked them.
For a product on this catalog it assembles a look for the product itself and a look for each of the ~200 products it finds similar to it, then groups them by style and returns up to four per style — sixteen to twenty looks, of which eleven to sixteen have a usable image for every item. Every piece in every look is the engine's choice, not ours; a look we could not
draw in full is dropped whole, never patched — and that rule cost us looks here. A board is the
engine's own authored layout, a fixed frame of slots keyed to garment category. No board in the library holds every combination of categories at once, so a look whose pieces cannot all be seated in one frame cannot be shown whole. The engine resolves that by dropping a piece from its own response; we will not show you a look with a garment quietly missing, so those looks are not on this rail at all.
Complete the outfit
bunsarcompleteTheLookthis product's own outfitsame emission — no second call
Classic High-Rise Straight Leg JeansbottomAdd
Leather Derby Dress ShoesshoesAdd
Multicolour Tote BagbagAdd
The product's own row, laid flat: the pieces the engine puts with
this sweater itself — one of its two Complete-the-Look boards, a second surface off the same
completeTheLook emission, with no second call.
Styled by
bunsarcuratedInspirationsthe same outfits, wornserved, not computed on the click
AI image · real catalog garments
the piece abovehero look
The Sweater, Worn
Every garment in this look, as its own product.
Mira & StoneStriped Wool Crewneck Sweatertop · $84.00Add
Atlas & CoClassic High-Rise Straight Leg Jeansbottom · $110.00Add
Lumen StudioStone Accent Metal Ringaccessory · $36.00Add
look #13573516
Drag the rail sideways: 11 looks, and every one of them opens — the picture large, and the look coming down garment by garment underneath. The first is the sweater from the page above, worn; the rest are the looks the engine built around pieces near it. Each look is one row the engine emitted for this catalog, untouched. The model imagery is AI-generated and marked as such on every frame; the garments inside it are real products from this catalog, and each one is the piece it actually named.
You may also like
bunsarmini-PLPbuilt from the outfits abovecomposed by us — not one engine feed
Lapel Collar Knit Sweatertop$74.00Slim Fit Mid-Rise Jeansbottom$43.00Suede Buckle Ballerina Flatsshoes$83.00Braided Pattern Tote Bagbag$95.00Leather Heart Stud Beltaccessory$24.00Striped Wool Crewneck Sweatertop$84.00Classic High-Rise Straight Leg Jeansbottom$110.00Leather Derby Dress Shoesshoes$183.00Multicolour Tote Bagbag$144.00Wool High Neck Sweatertop$84.00High-Rise Straight Leg Jeansbottom$49.00Structured Flap Shoulder Bagbag$195.00Suede Leather Belt With Metal Buckleaccessory$49.00Soft Cashmere Crewneck Jumpertop$87.00Slim Fit Capri Trousersbottom$82.00Structured Cropped Wool Coatouterwear$205.00Soft Gathered Loafer Shoesshoes$152.00Compact Handbag With Metallic Handlebag$94.00Wool High Neck Jumper With Belttop$39.00Wool Blend Ski Jumpertop$74.00High-Rise Barrel-Fit Jeansbottom$91.00Structured Shoulder Pad Coatouterwear$250.00Quilted Crossbody Handbagbag$180.00Argyle Pattern Knit Sweatertop$72.00
…the feed keeps going as long as the shopper keeps scrolling.
An endless product feed, specific to the piece they opened — this is the band
built to keep the shopper moving from one product to the next. It runs 24 deep here across 6 categories
(accessory, bag, bottom, outerwear, shoes, top). Read it honestly: every product in it is the engine's — each one is a
piece it styles with this garment, or with a garment it scored near this one. What is
ours is the decision to string those answers together into one feed, and the order they
come in. It emits the rows; it does not emit this list.
Similar pieces
bunsarsimilarProductsmore of the same category, endlesslyread from cache
The engine untouched: its own similarity feed for this product, in its own
order, nothing composed by us. It is deep, and it stays inside the product's own category — more knitwear, the jumpers the engine ranks nearest to this one. Nearness is not variety. Variety comes from the outfit bands higher
up — Style suggestions and Complete the outfit.
Personalisation, in one query parameter
Same page. Same second. Two shoppers.
The same product page re-orders itself around what each shopper is looking at — one query parameter, no profile.
GET /curatedProducts?retailerSKU=584444073&contextSKUs=…
Everything above was your shopper being dressed. This is the same
eye when your merchandiser opens it instead. Every figure below came out of the very
work you just scrolled through — nothing new was run to produce them. This is what it
already knew about your range, put into sentences.
“How much of my range can I actually sell as an outfit?”
883
products this catalog can style more than one way — its strongest anchors. Another 2,434 can be styled exactly one way. Two complete-the-look boards is the ceiling on this surface, so those 883 are where extra depth pays off.
measured · full count of its complete-the-look set, on the 4,999 products we publish
“And when it can't dress something?”
4
ways a piece ends up with no look — and the engine names which one it hit, for every piece: a class this setup doesn't dress, a piece it never reached for, a photo it couldn't read, or no photo at all. Which one it is tells you whose job it is — a setting, a buy, or a shot.
measured · our own demo catalog, full count, no sampling — every undressed piece bucketed by the outcome that stopped it
“Take one product. How many looks could I really put in front of a shopper?”
11–16
of the sixteen to twenty it builds around a single product. The rest exist and it can name every piece in them — they just have no photograph yet. This is a question about pictures, not about stock: Bunsar doesn't read a stock feed.
derived · five anchors, recomputed — not a catalog-wide average
That is the whole of what it does on this side. It reads the catalog it has
already indexed, it tells you plainly where the range is thin and where the pictures are
missing, and then it hands the work to a person. It doesn't go and change anything.
The questions are ours. The figures in the answers are its — each one comes out of the same emission that built the product page above. The admission that follows is ours, not the merchandiser's: of the outfits we could publish from this catalog at all, 14 lost a garment's photograph and were dropped whole rather than shown a piece short. That is our pipeline's limit, not a hole in anyone's range. We have not staged a session and called it a transcript.
bunsarthe team sidesame emission as the page abovederived — not a console session
What it can’t dress
Some pieces get no look. You get the reason — piece by piece.
There are four ways a garment ends up with no look. The engine knows
which one it hit, and it says so by name.
Not this setup. It read the garment. This configuration doesn’t build
looks around that class.
Never came up. Readable, allowed — the engine simply never reached for it.
Couldn’t read it. The photograph didn’t give it enough of the
garment to work with.
No image. There is nothing to read.
Which one it is decides whose job it is. A class you don’t dress is
a setting. A piece it never reached for is a buying question. A photograph it
can’t read is a photograph. Three different desks — and a percentage tells
you none of it.
And on the same shelf, a top it dresses twice.
Cotton Modal Strappy Top Tops
Look 1
TopsJacketsBags
Look 2
TopsTrousersShoes
Same catalog, same engine, same product group as pieces it
turns away. Nothing about this top is special. What differs is what the engine could
read and what the setup allowed — which is exactly what the reason names.
Point it at your own catalog and the list is yours: every piece it
can’t dress, each with the reason it didn’t.
measuredour own demo catalog, full count,
no sampling — every piece the engine builds into no look, bucketed by the outcome that
stopped it
The team side
Whoever's asking, it's the same engine underneath.
All three questions you just read came from one of these five — merchandising, wanting to know where the range is thin. The other four ask the same engine four different questions. They all get their answer from it, on their own catalog, in plain language.
Merchandising
Ask which pieces actually carry your outfits, and where the range has holes that leave a look unfinished.
Design
See how garments combine across the collection, and where a silhouette has nothing to pair with.
Commercial
Read the assortment the way it hangs together, not as a flat list of SKUs.
Campaign
Pull a themed edit — a palette, an occasion — assembled only from what is live in your catalog.
A GM view
One read across all of it, in a sentence, without waiting on a report.
The console
The room your team asks in. Here it is, rebuilt.
The three answers you just read live in a real surface — the one we built
and run. This is that surface itself, rebuilt for this page: the same frame, the same menu,
the same three module screens, holding only the figures this page has already shown its work
for. One voice answers; a fleet prepares behind it — and what it prepares lands back in the
same thread, then it hands the work to a person.
BUNSAR
Assistant /Fashion Intelligence⌂ HomeENTR
Search screens… ( / )
Dashboard▾
Recommended Actions
Assistant / Fashion Intelligence
File Library
Search▾
Visual Product Search
Product Management▾
Product DB Browser
Product DB Summary
Trend & Competitor▾
Trend Map 2D
Trend Map 3D
Attribute Trends
Color & Design▾
Similarity & Recommendation▾
Billing & Invoices▾
Modules▾
Modules / Your Plan
Merchandising
Pricing
Trend
Fashion IntelligenceYour assistant
How much of my range can I actually sell as an outfit?
883 products this catalog can style more than one way — its strongest anchors. Another 2,434 can be styled exactly one way. Two is the ceiling this engine built, so those 883 are where depth is worth buying.measured · full count of the complete-the-look set, on the 4,999 products we publish
And when it can’t dress something — do I just get a gap?
No. Every piece it can’t build a look around comes back named, with what stopped it: a class this setup doesn’t dress, a piece it never reached for, a photo it couldn’t read, or no photo at all. Which one it is tells you whose job it is — a setting, a buy, or a shot.measured · our own demo catalog, full count, no sampling — every undressed piece bucketed by the outcome that stopped it
Take one product. How many looks could I really put in front of a shopper?
That needs a fresh count rather than a guess. I’ll prepare it — the answer will land back in this conversation.
02:00Preparing…
11–16 of the sixteen to twenty it builds around a single product. The rest exist and it can name every piece in them — they just have no photograph yet. A question about pictures, not stock: no stock feed is read.derived · five anchors, recomputed — not a catalog-wide average
A task is being prepared in this
conversation. To keep the context clear, ask anything else in a new chat — this
panel unlocks itself the moment the answer is ready.+ New chat
Ask something…Send
History + New chat
Search conversations…
How much of my range can I actually
sell…this conversation — the one you are reading
Earlier conversations would sit here. We haven’t staged
any.
The countdown and the locked composer are the visible half of the loop:
when a question needs work the assistant hasn’t done yet, the fleet prepares it while
this conversation waits, and the answer lands back in the thread. Both normally remove
themselves at that moment — they are kept on screen here so you can see them.
MerchandisingPricingTrend
Merchandisingour own demo catalog
Your assortment health — breadth, depth, newness and coverage across
your own catalogue.
summaryEvery piece it can’t dress comes back with a reason
Not this setup. Read, and this configuration doesn’t build looks around that class.
Never came up. Readable, allowed — never reached for.
Couldn’t read it. The photo didn’t give it enough of the garment.
No image. Nothing to read.
Nothing lands in a gap unlabelled. The label is what routes the
work: a setting to change, a piece to buy around, or a photograph to re-shoot.
measured · full count, no sampling
suggestionDepth is worth buying on the 883
The 883 two-look anchors are the range’s
strongest pieces; the 2,434 one-look products are one garment away from a
second outfit.
measured · same count as the answer above
Pricingour own demo catalog
Your price position — bands, discount and markdown exposure across
your own catalogue.
summaryThe catalog trades between $19 and $258
$19$258
Every one of the 4,999 products carries a live price —
239 distinct price points across the band. The marks are the pieces styled on
this page, placed where they sit in it; no compare-at price exists in this feed, so no markdown story is claimed.
band: measured · marks: derived from the looks above
Trend is real and it is in the console — you can see its tab. It stays
shut on this page: its screen benchmarks against competitors, and we show nothing here we
haven’t measured on our own demo catalog. The one caveat it ships with is already on
this page: a benchmark to weigh, not a forecast to trust.
bunsarthe console, rebuiltour own demo catalogan illustration — not a screenshot, not a
session
Your stylists' judgment, on all 4,999 pieces
You style it. The engine scales it to the whole catalog.
Your stylists do the styling — that judgment is yours and it stays yours. Bunsar carries it, from one product to the whole catalog: point it at a single item and you get back not a "you may also like" row but several complete outfits, each garment set into a layout your team authored. The layout is a fixed frame. The intelligence is in the fill: it takes each empty slot and resolves it to a real product from the catalog — a piece that is visually right in that slot.
When a piece in an outfit has no usable photo, we drop the whole outfit rather than show it a piece short. Of the 42 outfits we could show at all here, 14 lost a garment's photograph and went in the bin whole; 28 survived, and those are the ones on this page. (It offered more: more than half of what it indexed on this catalog is real-brand material we never publish — the page runs on the 4,999 pieces we made ourselves, and those extra looks were set aside for that reason, not for a bad photograph.) What you see is therefore not a picture of a look. It is the look, rebuilt from the catalog itself — swap an anchor and the rest re-forms around it.
The engine's outfit around the Faux Suede Belted Jacketswap any piece below
The engine's outfit around the Woven Clutch Bag
The engine's outfit around the Hooded Faux Leather Jacket
Three outfits the engine built, on our own catalog — no brand's data on the page. Each one is a Complete-The-Look result: we point it at one product, and it completes the look around it. The three pieces it grouped with that product under one outfit are the three you see beside it. We chose which of its outfits to show, and where the pieces stand on the board; we did not choose the pieces.
What it resolved for the first look's slots
For every slot it also holds a pool of visually near pieces from the same catalog. One of them is in the first look above; the rest are what it would accept in its place. Choose any and the look re-forms around it.
outerwear
bottom
shoes
bag
Your data, and only yours
It runs on your data. We don't store it.
Bunsar works on top of your own catalog and your own engine. You pass it the pieces the shopper has been looking at, and it builds from there — then keeps none of it in our data model — no shopper profile, nothing logged.
Every brand runs on its own data, walled from every other. The eye gets sharper for everyone; the data never crosses. The capability is shared, never the data.
Structured Leather ToteFrom our own demo catalog
The index behind it
Your catalog is not the first one it has read.
We have been building and indexing at a scale that shows in how it reads a garment. These are totals across everything it has ever indexed since 2019. Most of it is the public assortments we have indexed, then the work we run in production, and last our own demo catalog, which is a rounding error inside the total. Each figure below is a combined total: no brand is broken out, and no brand's data is on this page. It is a record of what it has read, not a claim about what is on sale this morning.
19.8Mproduct records indexed since 2019
7.9Mpackshots read
3.4Mlooks
That is what stands behind a single filled slot.
The public assortments in that index are there for a reason: the same eye that reads a garment in your catalog can read one outside it, and place a piece of yours against its closest equivalents in the wider market. That belongs in a back office, not on a storefront — a benchmark to weigh, not a forecast to trust.
63 real cutouts from our own catalogOne screen of the index · the ledger beside it counts 19.8M · not a chart, not a result
What we don't put on this page
No number here is borrowed.
You have been shown conversion charts by three vendors already. We are not going to add a fourth.
We don't publish an uplift figure, because we don't take your sales feed. Any number we showed you would be someone else's store, dressed up as yours. The honest version is quieter: run it on your own catalog, and read the result there.
Everything you have seen on this page was built on our own catalog, so the capability is on the screen and no brand's data is. When you connect your data, the demo becomes your data — set up with our team, not a switch you flip alone.
See it on your own catalog.
Everything above ran on a catalog we made ourselves. The only way to judge it is to watch it read yours. Here is exactly what that costs you.
What we need from you
A product feed with images. That's the whole ask — no tracking script, no pixel, no tag on your site.
How long it takes
About two weeks to have it running on your own products. Three if your front-end team is starting from a blank page.
Do you have to touch your product page
Yes — and that is deliberate. We are an API company. You call us and place the results where you want them, styled the way your shop is styled. We do not inject anything into your page, and we cannot change how it looks. Your storefront stays yours.
Book a demo
NorthwindArgyle Pattern Knit Sweatertop · $72.00Add
Atlas & CoStraight Fit Trousers With Attached Beltbottom · $55.00Add
Lumen StudioWool Blend Longline Coatouterwear · $111.00Add
Lumen StudioLeather Ballet Flats With Embellishmentshoes · $66.00Add
AvenirDouble Layer Leather Belt With Star Detailsaccessory · $68.00Add
NorthwindClassic Crew Neck T-Shirttop · $57.00Add
AvenirRelaxed Fit Mid-Rise Jeansbottom · $98.00Add
Mira & StoneLeather-Effect Jacket With Attached Scarfouterwear · $149.00Add
AvenirLeather Loafer Shoesshoes · $178.00Add
Lumen StudioSplit Suede Maxi Tote Bagbag · $151.00Add
Mira & StoneSuede Leather Belt With Metal Buckleaccessory · $49.00Add
NorthwindAsymmetrical Knit Midi Skirtbottom · $62.00Add
Lumen StudioFluid Longline Blazer With Pocketsouterwear · $191.00Add
Atlas & CoStrappy Cut-Out Ballet Flatsshoes · $160.00Add
Lumen StudioStructured Shoulder Bag With Metal Accentsbag · $181.00Add
Mira & StoneGeometric Buckle Leather Beltaccessory · $49.00Add
Lumen StudioRibbed Crop Toptop · $81.00Add
Atlas & CoTextured Mini Skirt With Contrast Detailsbottom · $105.00Add
NorthwindSatin Ballet Flats For Sportsshoes · $92.00Add
Lumen StudioNatural Jute Tote Bagbag · $131.00Add
Lumen StudioAdjustable Cord Necklace With Stone Accentsaccessory · $51.00Add
AvenirClassic Denim Button-Up Shirttop · $38.00Add
NorthwindStructured Tuxedo-Style Trousersbottom · $87.00Add
Atlas & CoGathered Leather Loafer Shoesshoes · $90.00Add
AvenirMini Beaded Crossbody Bagbag · $128.00Add
NorthwindFloral Embellished Leather Effect Beltaccessory · $57.00Add
Mira & StoneContrast Panel Shortsbottom · $49.00Add
Lumen StudioClassic Check Pattern Shirttop · $56.00Add
Atlas & CoSatin Bow Ballet Flatsshoes · $70.00Add
AvenirBraided Oversized Tote Bagbag · $173.00Add
Atlas & CoResin Disc Adjustable Beltaccessory · $50.00Add
AvenirHigh-Rise Wide-Leg Denim Jeansbottom · $63.00Add
AvenirRuffled Short Sleeve Toptop · $88.00Add
NorthwindSoft Gathered Loafer Shoesshoes · $152.00Add
Atlas & CoStructured Flap Shoulder Bagbag · $195.00Add
Atlas & CoHigh-Rise Tailored Trousersbottom · $55.00Add
Lumen StudioLeather Platform Loafer Shoesshoes · $76.00Add
Mira & StoneShimmer Weave Crossbody Bagbag · $169.00Add
AvenirResin Bead Statement Necklaceaccessory · $23.00Add
Lumen StudioRuched Sleeveless Toptop · $31.00Add
Atlas & CoStraight Fit Trousers With Attached Beltbottom · $55.00Add
Lumen StudioDenim Shirt With Bow Detailtop · $56.00Add
NorthwindDistressed Loose-Fit Jeans With Lace Detailbottom · $72.00Add
Atlas & CoPolka Dot Ballet Flats With Bow Detailshoes · $75.00Add
AvenirStructured Crossbody Bagbag · $163.00Add
Lumen StudioStone Accent Metal Ringaccessory · $36.00Add
