Series DTrax

Trax puts computer vision on every retail shelf

Singapore-headquartered Trax turns ordinary shelf photos into SKU-level analytics for consumer-goods giants like Coca-Cola and Nestle — a retail-AI unicorn since 2019 whose image-recognition arm merged with FORM in 2026.

منصة سنغافورية للرؤية الحاسوبية تحوّل صور رفوف المتاجر إلى تحليلات فورية للمخزون وتوافر السلع، وتخدم علامات استهلاكية عالمية مثل كوكاكولا ونستله.

Next →
  1. The Bottleneck

    What was broken?

    Brands and retailers had no reliable, real-time view of what was actually on store shelves — out-of-stocks, misplaced products and poor displays cost sales, and manual audits were slow, expensive and unscalable.

  2. The Move

    Why it worked

    Applied computer vision to a concrete, high-value job — photograph a shelf, get back SKU-level availability and merchandising insights — and sold it to the world's largest CPG companies, scaling to 150+ enterprise customers from a Singapore HQ.

  3. The Trap

    Battle scars

    Enterprise retail sales cycles are long and lumpiness in big contracts hurts; shelf-AI rivals contest every deal; and the 2025 carve-out of the image-recognition business to Gemspring Capital plus the 2026 FORM merger show how much restructuring the business needed to keep growing.

The problem

For a brand like Coca-Cola or Nestle, the shelf is the moment of truth — but headquarters could not see it. Manual audits covered a fraction of stores and arrived weeks late, so out-of-stocks, missing promotions and broken planograms silently ate sales every day.

How it works

Trax sends computer vision down the store aisle. Photos taken by field reps or fixed cameras are processed into SKU-level reads: what is in stock, what share of shelf each brand holds, whether the agreed display is actually up, and at what price. Dashboards turn that into corrective tasks while acquisitions — Qopius for real-time shelf monitoring, Planorama for European image recognition, Shopkick for shopper rewards data, Nielsen’s store-observation assets — widened the data moat.

Pain points

Empty shelves nobody reports, promotional displays that never get built, planogram compliance near zero in far-flung outlets, and audit reports so stale they describe a store that no longer exists.

Business model

Enterprise SaaS: annual subscriptions and analytics licenses per brand or retailer, complemented by project-based in-store data collection and activation products that monetize the shopper side.

Challenges

Long enterprise sales cycles with lumpy mega-contracts, direct rivalry from other shelf-AI vendors, and the strategic cost of breadth: in 2025 Trax divested its image-recognition business to Gemspring Capital, and in 2026 that unit merged with FORM — a reminder that even category pioneers must restructure to keep scaling.

Funding

  • Raised: $100M Series D led by HOPU Investments (Jul 2019) at a valuation over $1.1B, making Trax Singapore’s second unicorn; total equity financing topped $350M with Warburg Pincus as largest shareholder alongside Boyu Capital, Investec and GIC (Retail Technology Innovation Hub).
  • Valuation: over $1.1B at Series D (Jul 2019).

Latest — February 2026

Gemspring Capital merged FORM (GoSpotCheck, FORM OpX) with Trax’s image-recognition business to create a combined retail-execution and shelf-intelligence platform serving 750+ customers worldwide under FORM CEO Ali Moosani (PR Newswire via Third News) — the transaction that repositioned Trax Ltd.’s remaining businesses toward growth and market expansion.

Similar sparks

⚡️ SERIES EVal: over $1B (Series E, Mar 2021)

Singapore · 2026-06-16

PatSnap turns the world's patents into R&D intelligence

Singapore-born PatSnap mines 120M+ patents, papers and grants with AI so R&D and IP teams can track rivals, vet ideas and find partners — a $1B+ unicorn now lining up a dual Hong Kong–Singapore IPO.

بالعربي

منصة سنغافورية تبحث في أكثر من 120 مليون براءة اختراع بالذكاء الاصطناعي لمساعدة فرق البحث والتطوير على تتبّع المنافسين واكتشاف فرص الابتكار.

  • AI/SaaS
  • #singapore
  • #ai
⚡️ PRE-SERIES A$16.5M total (incl. $6.5M pre-Series A)

Syria · 2026-09-26

aiXplain lets teams build AI solutions without writing code

A no-code AI marketplace where businesses assemble ready-made models into custom solutions and action-taking agents, benchmark them side by side, and deploy paying as they go — with a focus on Arabic-language representation.

بالعربي

منصة ذكاء اصطناعي بلا برمجة تتيح للشركات بناء الحلول والوكلاء الأذكياء ومقارنتها ونشرها مع تمثيل أفضل للغة العربية.

  • AI/SaaS
  • #syria
  • #ai
⚡️ SERIES B+Val: $2.65B

Japan · 2025-11-17

Sakana AI breeds small, efficient Japanese-first models

Build affordable generative-AI models tuned for Japanese language and small datasets, using evolutionary methods instead of brute-force scale.

بالعربي

طوّر نماذج ذكاء اصطناعي توليدي ميسورة التكلفة مضبوطة للغة اليابانية والبيانات الصغيرة بدل سباق الحجم المكلف.

  • AI/SaaS
  • #japan
  • #ai
⚡️Spark playbook copied! Go build.