Home/Fynd Horizon
Recent Innovations · Ultra Fast Retail · Made-to-Order in 24 hours

Fynd Horizon.

The same Trends store, 25× the SKUs, made-to-measure in 24 hours — body scan in three clicks, virtual try-on in under two seconds, doorstep in 30 minutes or 24-48 hours.

Live · RCP cluster · 31-Mar-2026 Live · Mumbai cluster · 27-Apr-2026 Live · Tirupur MTO factory · cycle running since 25-Mar-2026
Headline numbers · 27-Apr-2026
SKUs available · per store
50,000+
all from JCP unified catalog
vs Trends today
~2,000
SKUs displayed per store today
Body-measurement accuracy
98.3%
Ratl study · n=200 · 18-Apr-2026
Order → doorstep
30 min / 24-48 hr
standard / made-to-order
Fynd Horizon LED-wall installation at a Reliance Trends store · Infinite Aisle catalog grid + branded sign
Live installation · Reliance Trends store · LED wall renders Infinite Aisle catalog + life-size customer try-on. Photo: in-store, 27-Apr-2026.
01 · Status · 27-Apr-2026

What's running. What's being built. What's next.

Running today
Live
  • RCP store · live to customers since 31-Mar-2026 · first Horizon-enabled Trends store
  • Mumbai cluster · live 27-Apr-2026 · LED wall + iPad + RealSense kit deployed in-store
  • Tirupur partner factory · 24-hour MTO cycle running since 25-Mar-2026 · automated cut + sew + AI QC
  • Infinite Studio admin · operator surface live · 178 try-ons in week of 24-Apr-2026 → 01-May-2026 · 103 looks generated · 25% engagement
  • 14-Apr-2026 · MDA + MM Sir unveil at RCP · 30-min deep engagement
  • Catalog from Fynd UNO · Rendering QA by Ratl · Order routing through JCP · the three-platform substrate
Building this quarter
Building
  • Store-by-store rollout · expanding across the Mumbai cluster after the 27-Apr go-live
  • Demand-prediction agent · next-3-outfit forecast feeding tonight's pre-cut at the Tirupur factory
  • Token-economics dashboard · per-session cost vs AOV tracked daily — break-even at AOV ≥ ₹780
  • Self-healing rendering · Ratl QA-agent reviewing 100% of nightly renders · drift-flag pipeline hardening
Next horizon
Roadmap
  • 15-Jul-2026 · public launch · Nexus Seawoods
  • Pre-cut pipeline at scale · target factory yield +18% by Q2 2026
  • Expansion beyond Mumbai cluster · sequencing TBD with Trends ops
  • On-call rotation · 5-minute page-out on any photoreal drift > 0.5%
  • L3 autonomy targets · per the autonomy framework that anchors the 30-Apr-2026 letter to MM Sir
02 · The strategic shift

From inventory-heavy retail to AI-native ultra fast retail.

Trends today is constrained by floor square footage — when a customer's size or colour isn't on the floor, they walk out. Horizon decouples the store from physical stock: scan, try, buy, manufacture, deliver — without ever holding inventory.

LensTrends todayHorizon
SKUs per store~2,000 displayed · physical only50,000+ · all virtual via JCP catalog
Inventory riskMarkdown risk on every garment · ~45 days holdingZero · only manufacture what's sold · 0 days holding
Returns rate8.5% online · sizing-driven<2% target · deep-tech fit eliminates sizing returns
Order → doorstep48 hr (standard) · 7 days (custom)30 min (dark store) · 24-48 hr (MTO factory)
Revenue densityLimited by floor sq ft25× reach from same footprint

Strategic shift: Ultra Fast Retail powered by automated factories that turn orders into products in 24 hours. Deep-tech cameras for precision measurement → automated factory for 24-hour manufacturing → next-day logistics for rapid fulfilment.

03 · User journey · 6 steps

One store. One system. Infinite inventory.

The end-to-end customer flow at any Horizon-enabled Trends store. Step 0 is one-time onboarding; steps 1-5 repeat per visit.

Step 0 · Body Scan
Live

3-click profile.

Selfie + front-full + side-full. Captures shoulder width, chest, waist, inseam · plus depth, posture, and volume for 3D reconstruction.

Deep tech: Intel RealSense depth cameras · SMPL parametric model infers body shape under clothing · 98.3% accuracy (Ratl study, n=200).
Step 1 · Browse
Live

50,000+ styles at your fingertips.

Swipe through the entire brand inventory · filter by occasion, colour, trend, or price · single-tap to add to the Virtual Try-On queue.

Deep tech: Catalog from Fynd UNO · live sync with brand inventory.
Step 2 · Try
Live

See it on you, not on a model.

Visualise any garment on your own body · best-fit size recommendation from your scan data · mix & match complete outfits in seconds.

Deep tech: Gemini Cloud Rendering · physics-accurate fabric drape · <2 s render · generative-AI lighting and texture.
Step 3 · Buy
Live

One-tap checkout. "Manufacturing started."

One-tap purchase of items just tried on virtually · instant confirmation · live track from "Order Placed" → "Cutting Fabric".

Deep tech: Unified OMS routes to optimal facility · factory API integration · real-time fabric availability check.
Step 4 · Make
Building · scaling

Made just for you. 24-hour turnaround.

Garment manufactured from scratch — not pulled from a shelf. From raw fabric to finished product in less than a day. Every stitch matches the customer's exact specifications.

Deep tech: Automated laser cutting · robotic sewing lines · AI computer-vision defect inspection on every seam.
Step 5 · Deliver
Live

30 min or 24-48 hours. Doorstep.

Standard fit from local Trends dark store in 30 min · returns OK within 7 days. Custom MTO from Tirupur in 24-48 hr · remake-free guarantee if fit imperfect.

Deep tech: AI route optimisation · automated sortation · predictive logistics for courier capacity.
Trends Infinite Aisle catalog grid · 14 garment thumbnails (tops + bottoms + dresses) shown to a customer browsing in-store
Step 1 evidence · Infinite Aisle catalog grid

What the customer sees on the iPad and the LED wall during Browse. Catalog rendered from Fynd UNO · 50,000+ SKUs paged through filter and search.

Catalog from Fynd UNO · Rendering QA by Ratl · Order routing through JCP

04 · In-store hardware kit

Plug-and-play. ≤6 hours per store.

Standardised SKUs · zero local config · the same kit ships to every Horizon-enabled Trends store. Five components — one wall, one tablet, two depth cameras, one edge box, one webcam — assembled in under a working day.

Component 1 · Display

LED Wall

16 ft × 8 ft 4K LED wall · life-size mirror surface for the virtual try-on render.

Component 2 · Customer + staff control

iPad Pro 12.9″

Catalog browse · checkout · staff control surface for manual override.

Component 3 · Depth sensing

Intel RealSense D555

Dual depth cameras · captures the 3D body scan that powers measurement and try-on.

Component 4 · Local compute

Edge Compute Box

Local inference fallback when the network goes down · 90-second cold-start warmup.

Component 5 · Capture

48 MP WebCam

Full-body picture for the customer profile and the try-on composite.

Render of the in-store hardware setup · LED wall + iPad kiosk + customer in front of the display
05 · AI-native layer

Predicting the future. Not reflecting the past.

Three agents make Horizon AI-native rather than AI-flavoured: a demand-prediction loop that pre-cuts factory fabric tonight for tomorrow's predicted orders, a self-healing render-QA loop that catches model drift before customers see it, and a token-economics loop that meters every try-on against a profitability threshold.

Agent 1
Building

Demand prediction

Predicts the customer's next 3 outfits from scan + browse history. Surfaces top-12 looks before the customer asks. Pre-cuts factory fabric tonight for tomorrow's predicted orders.

Target: factory yield +18% by Q2 2026.
Agent 2
Live

Self-healing rendering

Auto-reviews 100% of VTO renderings every night via the Ratl QA agent. Flags photoreal drift > 0.5% to the DRI within the hour. Catches Gemini-cloud regressions before customer impact.

SLA: on-call rotation pages within 5 minutes of a red flag.
Agent 3
Live

Token economics

Every try-on metered at ₹4.20 / session today. Profit threshold: AOV must clear ₹780 to break even. Hard cap: 50 try-ons per customer per session.

Outcome: per-session profit-positive at every store.
Infinite Studio admin dashboard · Weekly Performance Report showing 178 try-ons, 17 today, 58% completion rate, 103 looks, 25% engagement, ₹1358 estimated profit, 14-day activity trend, key highlights, peak hours, size distribution, recent sessions
Operator surface · Infinite Studio admin · Weekly Performance Report (24-Apr-2026 → 01-May-2026)
Live

178 try-ons · 17 today · 58% completion rate · 103 looks generated · 25% engagement · ₹1,358 estimated profit. The metering loop in Agent 3 isn't aspirational — it's the surface that operators use today. 14-day activity trend, peak hours, size distribution, and per-session detail all live.

06 · Two fit paths

Speed or measure-perfect. Both end with a perfect fit.

Re-framed from D01: not casual vs formal — speed vs measure-perfect. The customer picks the trade-off they want at checkout.

Path A · Standard Fit
Live

Ready to wear.

30 min · doorstep
  • · Best-matched standard size from scan (S / M / L / XL)
  • · Fulfilled from local Trends dark store (Mumbai cluster · 3 stores)
  • · Inventory pre-positioned · no-stockout > 99% of the time
  • · Free returns within 7 days · max 2 per customer (MDA policy)
Path B · Measure-Perfect Fit · Next Gen
Building · scaling

Made to order.

24-48 hr · doorstep
  • · Manufactured to exact body measurements from depth scan
  • · Tirupur partner factory · automated cut + sew + AI QC
  • · Pre-cut fabric tonight for tomorrow's predicted orders
  • · Remake-free guarantee if fit imperfect · owner: factory DRI
07 · Benchmarks · vs Uniqlo, Shein, Zara

Benchmarked against the global standard. Not against ourselves.

The honest read on how Horizon compares to the fast-fashion playbooks the world has already validated. Targets are Ratl-verified — not internal projections.

MetricHorizon targetTrends todayUniqlo TokyoShein on-demandZara
SKU reach / store50,0002,000~8,000600,000 (online only)~10,000
Returns rate<2%8.5% (online)~6%~12%~10%
Order → doorstep · custom24-48 hr7 daysn/a (no MTO)5-10 days14 days
Order → doorstep · standard30 min48 hrn/an/an/a
Inventory holding0 days (MTO)45 days~14 days~3 days14 days
08 · Research · Mathematics behind measurements

A single phone camera + a 100-year-old equation = tape-measure-grade body data.

The CEO Brief on the math powering Horizon's body-measurement engine. Three pillars: 17-keypoint pose estimation, depth-sensing pixel→world projection, and Ramanujan's ellipse approximation for body circumferences. Read it inline below or download the 2-page PDF.

Pillar 1

Pose Estimation · 17 Keypoints

AI vision model maps the human body to 17 anatomical landmarks (head/face · upper body · lower body) in real time from a single camera frame. Output: 2D pixel coordinates + confidence per joint.

Pillar 2

Depth Sensing · Pixel → Real World

Depth map gives Z-distance for every pixel. Converts a flat image into a 3D point cloud. Drives Euclidean distances between joints (height, arm, inseam, shoulder width, torso length).

Pillar 3

Ramanujan's Ellipse · Circumferences

Body cross-sections (waist, chest, hips, thigh) modelled as ellipses. Ramanujan's Approximation 3 gives near-exact perimeter — error < 0.04% across all eccentricities · runs on-device in microseconds.

Accuracy
± 1 cm measurement · < 0.04% formula error · 98.3% body-shape accuracy vs Bodygram (n=200, Ratl-verified)
Download PDF · 1 pp · 410 KB ↗