CHALLENGE · 01
Legacy monolith won't scale to peak
Magento or Salesforce Commerce Cloud monolith. Black Friday breaks the same thing every year. The team dreads November.
Headless · composableIndustry · 05 of 06 · DTC · Marketplaces · Omni-channel
The store that knows the shopper on their second click.
Personalization isn't a banner — it's a stack. Real-time identity, search relevance, demand forecasting and the OMS plumbing that keeps the "in stock" badge honest at midnight on Black Friday.
+24%Conversion lift
7×Paid-media ROAS · best cohort
< 150msSearch · p95
storefront · live · 14:02 · IN/EUROAS 6.8×
Sessionslast 60 min
14,820
Product viewrec hit
10,04167.7%
Add to cartreal-time price-fit
4,74232.0%
Checkoutguest + wallets
3,26522.0%
Purchase+24% vs. legacy
2,22815.0%
For Riya · returning · session 4
Linen shirt0.94
Slip dress0.91
Tote bag0.86
The reality on the ground
What's actually moving the needle in commerce.
Not the redesign. Not the new logo. The boring infrastructure: identity resolution, real-time inventory, search that respects margin, and an OMS that doesn't lie about "in stock" on a marketplace at 11pm.
FACT · 0170%
Average cart-abandonment rate. Half of it traces back to friction the team has the data to remove — and hasn't.
FACT · 02$5.9T
Global e-commerce in 2026. The unit economics for the average DTC brand are getting worse, not better. Performance marketing isn't a moat.
FACT · 039×
Peak vs. average traffic on Black Friday. The legacy commerce stack was sized for the average. Nobody believes it on the night.
FACT · 044.2
Average number of channels per shopper journey. The brand that owns identity across them owns the customer.
The hard problems
What keeps a CTO & Head of Growth awake at 02:00.
Commerce problems compound. A slow search drops conversion, drops AOV, drops LTV — and the CFO blames acquisition cost. We focus on the upstream defects.
CHALLENGE · 02
Personalization that ranks by margin
Most rec engines optimize for click. Yours should optimize for contribution margin — and respect what's actually in stock.
Real-time + marginCHALLENGE · 03
Search that lies about relevance
"Linen shirt" should return a linen shirt — not a polyester one with "linen-look" in the description.
Vector + lexical hybridCHALLENGE · 04
Inventory truth across channels
Web says yes, store says no, marketplace listed it an hour ago. Customer is annoyed. Cancellation rate climbs.
Distributed OMSCHALLENGE · 05
Demand forecasting that holds up
SKU-store-day forecasts that work for fashion, fresh and long-tail — not just the top 100 SKUs.
Hierarchical MLCHALLENGE · 06
Identity across web · app · store
CDP that resolves the same shopper across guest checkouts, store visits and marketplace orders — without violating consent.
CDP + consentWhat we build · and own
Six modules. One commerce spine.
Headless or hybrid, replatform or augment — we work with the stack you already have, ship the modules that move the needle first, and replatform only what truly needs it.
Module · 01
Headless & composable commerce
Storefront in Hydrogen or Next, commerce in Shopify or commercetools, content in Sanity or Contentful. Replatform without freezing the roadmap for nine months.
- Next.js / Hydrogen
- commercetools / Shopify
- Sanity · Contentful
- Vercel · Edge networks
Module · 02
Real-time personalization
Identity-stitched event stream into vector + lexical hybrid search. Ranking that respects margin, recommendations that respect inventory, prices that respect contracts.
- Real-time CDP join
- Margin-aware ranking
- Inventory-aware recs
- Personalized landing
Module · 03
Search & relevance
Sub-150ms search with hybrid retrieval, typo tolerance, semantic boosters, merchant rules. Faceting that actually narrows the result set instead of clearing it.
- Vector + BM25 hybrid
- Synonym + typo handling
- Merchant rule overrides
- A/B + bandit testing
Module · 04
Demand forecasting & OMS
SKU-store-day forecasts you can argue with. Order routing across stores, DCs and dropship partners. "In stock" that means in stock, even on the sixth marketplace.
- Hierarchical forecasts
- Lifecycle-aware ML
- Manhattan · IBM Sterling
- Distributed inventory ledger
Module · 05
Loyalty · CDP · customer 360
A real customer-360 view, not a slide. Loyalty mechanics tuned to LTV. Lifecycle journeys that respect consent and the GDPR right-to-be-forgotten.
- Segment · mParticle · Lytics
- Loyalty engine
- Lifecycle journey orchestration
- Consent + DSAR APIs
Module · 06
Checkout & payments
Frictionless guest checkout, saved wallets, regional rails (UPI / Pix / iDEAL), 3DS2 fraud, and a post-purchase that doesn't lose the customer.
- Apple/Google Pay · Link
- 3DS2 step-up
- PSP-orchestrated fallback
- Post-purchase tracking
Reference architecture
From first session to repeat purchase, the path is six steps.
From the first visit to the seventh order, every shopper event flows through six clean stops — measured, optimized, never re-collected.
STEP · 01CaptureWeb / app / store events streamed to the CDP in real time.
STEP · 02ResolveIdentity stitched · consent applied · profile fan-out.
STEP · 03RankRecs + search ranked on margin · inventory · LTV signal.
STEP · 04ServeEdge-rendered storefront · sub-150ms · personalized at first paint.
STEP · 05FulfilOMS routes order across stores, DCs, dropship partners.
STEP · 06LearnOutcomes feed back into forecasts · ranking · journeys.
Tools we reach for
Battle-tested stack. No buzzword tax.
Storefront & content
Next.js · App RouterHydrogen · RemixSanity · ContentfulVercel · Edge FunctionsTypeScriptTailwind
Commerce platforms
commercetoolsShopify PlusBigCommerceStripe · AdyenMagento AdobeSAP CX
Data & personalization
Segment · mParticleAlgolia · TypesenseBigQuerySnowflakeIcebergHightouch
OMS · fulfillment
Manhattan ActiveIBM Sterling OMSfluent commerceShipstationEasyPostVeeqo
Where we've shipped in commerce.
From DTC start-ups to omni-channel retailers, marketplaces and big-box brands going direct — and the quick-commerce teams running 10-minute SLAs.
DTC brands · D2CFashion · beauty · luxuryBig-box · multi-bannerGrocery & quick commerceMarketplaces · B2B/B2CSubscription commerceComposable migrationsHeadless replatforms
Case study · in production
How a DTC fashion brand replatformed in 14 weeks — and survived Black Friday.
A monolithic Magento that broke on every peak. Six markets, four currencies, three loyalty programmes. We replatformed onto commercetools + Hydrogen, layered Algolia hybrid search and a real-time CDP. Black Friday peaked at 9× normal RPS, and the on-call channel was quiet.
ClientDTC fashion · 6 markets
GMV$140M · annualized
Timeline14 weeks · replatform live
Peak9× normal RPS on Black Friday
+24%Conversion lift · YoY
7×Paid-media ROAS · best cohort
< 150msSearch · p95
0 incidentsBlack Friday · weekend
Compliance · built-in
Audit-ready by default. Not by sprint.
Card data, shopper PII, consent across regions. We don't take shortcuts on any of them — your CFO doesn't need a card-data breach in their Q4.
PCI-DSS · SAQ-ATokenized vault · no PAN
SOC 2 · Type IIContinuous controls
GDPR · CCPA · DPDPMulti-region consent · DSAR
ADA · WCAG 2.2 AAAccessibility · litigation-grade
PSD2 · 3DS2EU strong-customer auth
OWASP ASVS L2Application security baseline
ISO 27001Information security
CSA STAR · cloudCloud security · level 2
Questions you’ll probably ask
The short version of the kick-off call.
Do we have to replatform to get value?+
How do you handle Black Friday / peak?+
How long does a typical engagement run?+
How do you measure personalization without spurious uplift?+
Can you integrate with our existing ERP / OMS / WMS?+
How do you handle consent and privacy across markets?+
Commerce · DTC · marketplaces
Share your funnel. We'll show you the upstream defects.
A 60-minute working session with a senior engineer and a commerce lead. We'll review your stack, your peak-event history and the three places we'd move first.