Solutions. Custom systems behind the storefront: routing, inventory, checkout, B2B, returns, and AI that keep ops and conversion in sync.
Engineering solutions
Nearest-farm order routing.
When an order lands, the system scores nearby farms on distance, inventory, cut time and capacity, then routes fulfillment to the grower that can ship the freshest product. Cut-offs, blackout windows and partial-fulfillment rules stay in one place, so marketing can sell nationwide without ops hand-picking a warehouse on every order.
Nearest-farm order routing.
When an order lands, the system scores nearby farms on distance, inventory, cut time and capacity, then routes fulfillment to the grower that can ship the freshest product. Cut-offs, blackout windows and partial-fulfillment rules stay in one place, so marketing can sell nationwide without ops hand-picking a warehouse on every order.
The problem
Perishable product shipped from a default location arrives late, and ops end up re-assigning orders by hand every morning.
What we build
- A routing service that scores every eligible location the moment an order is created, then moves the fulfillment order to the winner.
- Cut-off times, blackout dates and daily capacity stored in metaobjects, so ops can change them without a deploy.
- Split and fallback rules for partial stock, with every routing decision logged on the order timeline.
Talk to an engineerStack
Returns-to-refurb loop.
Returned units move through a grading queue (resell, refurb, scrap) with disposition rules tied to condition and SKU economics. Sellable inventory is written back into a certified pre-owned or outlet channel with the right price, photos and warranty flags, so circular stock doesn’t collide with new-product availability.
Returns-to-refurb loop.
Returned units move through a grading queue (resell, refurb, scrap) with disposition rules tied to condition and SKU economics. Sellable inventory is written back into a certified pre-owned or outlet channel with the right price, photos and warranty flags, so circular stock doesn’t collide with new-product availability.
The problem
Returns pile up in a corner of the warehouse. Good stock gets written off because nobody has a fast way to grade it and sell it again.
What we build
- A grading queue for returned units, with resell, refurb and scrap outcomes driven by condition and SKU margin.
- Graded stock written back as certified pre-owned or outlet products, with their own price, photos and warranty metafields.
- A separate location and channel for circular stock, so it never inflates new-product availability.
Talk to an engineerStack
Post-purchase upsell that respects ops.
After checkout, the offer layer only surfaces add-ons while the order can still be amended in the warehouse window. Accepted upsells patch the existing order and the fulfillment feed, so revenue doesn’t create a second shipment or a pick-list mismatch.
Post-purchase upsell that respects ops.
After checkout, the offer layer only surfaces add-ons while the order can still be amended in the warehouse window. Accepted upsells patch the existing order and the fulfillment feed, so revenue doesn’t create a second shipment or a pick-list mismatch.
The problem
Post-purchase offers lift revenue, but each accepted one creates a second shipment and a pick list the warehouse didn’t expect.
What we build
- Thank-you and order status page offers built as checkout UI extensions.
- Eligibility checks against the warehouse release window, so offers disappear once an order is picked.
- Accepted offers added to the existing order through order editing, not as a new order.
Talk to an engineerStack
Launch-day traffic shed.
Drops and launches stay up when traffic spikes. Pages are built to cache, live stock loads on its own, and checkout rules cap quantities per customer and turn away bots, so a sell-out moment reaches real fans instead of resellers.
Launch-day traffic shed.
Drops and launches stay up when traffic spikes. Pages are built to cache, live stock loads on its own, and checkout rules cap quantities per customer and turn away bots, so a sell-out moment reaches real fans instead of resellers.
The problem
Limited drops bring a wave of traffic and bots in the first minutes. Pages slow down, stock oversells and real customers miss out.
What we build
- Drop pages and product pages trimmed to what the CDN can cache, with live stock loaded separately.
- Per-customer purchase limits and bot rules enforced in checkout with Shopify Functions.
- A launch runbook: load test, rollback plan and a war-room dashboard for the first hour.
Talk to an engineerStack
AI packing-slip exception triage.
Before an order is picked, an AI pass reads the notes, gift messages and address and flags anything a person should see. Ops gets a short queue of exceptions with a suggested fix, while clean orders flow to the warehouse untouched.
AI packing-slip exception triage.
Before an order is picked, an AI pass reads the notes, gift messages and address and flags anything a person should see. Ops gets a short queue of exceptions with a suggested fix, while clean orders flow to the warehouse untouched.
The problem
Odd orders slip through to the warehouse and get caught at the packing bench, or not at all, and turn into support tickets.
What we build
- A model that reads order notes, gift messages and addresses before an order is released to pick.
- Exceptions sorted into a queue with a suggested fix, such as freight to a PO box or a gift note that won’t fit the slip.
- Clean orders pass straight through, so the team only ever sees the few that need a person.
Talk to an engineerStack
Multi-warehouse inventory.
One source of truth for stock across warehouses, 3PLs and retail. Inventory syncs both ways with the ERP, buffers protect each channel, and shoppers see availability and delivery dates for their region instead of a single global number.
Multi-warehouse inventory.
One source of truth for stock across warehouses, 3PLs and retail. Inventory syncs both ways with the ERP, buffers protect each channel, and shoppers see availability and delivery dates for their region instead of a single global number.
The problem
Stock lives in an ERP, a 3PL and Shopify at once, and the three numbers rarely agree.
What we build
- Two-way sync between the ERP or WMS and Shopify locations, with a full audit trail.
- Safety stock per channel and location, so retail, wholesale and DTC can’t sell the same unit twice.
- Region-aware availability and delivery dates on the product page.
Talk to an engineerStack
B2B pricing and net terms.
Wholesale buyers log in to the same storefront and see their own catalog, pricing and net terms. Price lists and company accounts sync from the ERP, so sales and finance stop maintaining a second store.
B2B pricing and net terms.
Wholesale buyers log in to the same storefront and see their own catalog, pricing and net terms. Price lists and company accounts sync from the ERP, so sales and finance stop maintaining a second store.
The problem
Wholesale runs on a separate store, a PDF price sheet and email orders, so it never gets the attention the DTC site does.
What we build
- Shopify B2B companies, locations and catalogs synced from the ERP.
- Customer-specific price lists, volume breaks and payment terms applied at checkout.
- A single storefront and theme for retail and trade buyers, so there’s one codebase to maintain.
Talk to an engineerStack
Bulk order pad and quick reorder.
Built for buyers who know what they want. A bulk order pad, matrix entry and CSV upload get a full order into the cart in minutes, and reorders take one click. Case packs and minimums are checked before checkout, not after.
Bulk order pad and quick reorder.
Built for buyers who know what they want. A bulk order pad, matrix entry and CSV upload get a full order into the cart in minutes, and reorders take one click. Case packs and minimums are checked before checkout, not after.
The problem
Trade buyers order dozens of SKUs at a time. Clicking through product pages one by one sends them back to email.
What we build
- A quick-order pad with SKU search, a size and colour matrix, and CSV upload.
- One-click reorder from past orders and saved lists.
- Minimums, increments and case-pack rules enforced in the cart.
Talk to an engineerStack
Bundles that ship as components.
Sell kits and bundles without a phantom SKU. The shopper sees one product and one price, while inventory and the warehouse see the real components, so stock stays accurate and the right items go in the box.
Bundles that ship as components.
Sell kits and bundles without a phantom SKU. The shopper sees one product and one price, while inventory and the warehouse see the real components, so stock stays accurate and the right items go in the box.
The problem
Bundles were set up as their own SKUs, so stock went out of sync with the components and the warehouse picked the wrong items.
What we build
- Bundles built with a cart transform Function, so each one is a single line for the shopper.
- Component SKUs passed to the 3PL and decremented from inventory.
- Bundle pricing and savings shown on the product page and in the cart.
Talk to an engineerStack
Product variant architecture.
How options, colourways and SKUs are modelled on Shopify Plus, so inventory, product pages and merchandising stay in sync without a brittle option matrix. One clean model feeds swatches, filters and feeds alike.
Product variant architecture.
How options, colourways and SKUs are modelled on Shopify Plus, so inventory, product pages and merchandising stay in sync without a brittle option matrix. One clean model feeds swatches, filters and feeds alike.
The problem
Catalogs outgrow the default option matrix. Merch teams end up with duplicate products, broken swatches and filters that don’t match.
What we build
- Product models mapped out before build: options, colourways, sizes and SKUs.
- Combined listings and metaobject-driven swatches for large ranges.
- Collection and search filters that read from the same structured data.
Talk to an engineerStack
Checkout rules that prevent bad orders.
Business rules enforced at checkout, not in the support inbox. Restricted products, PO boxes, hazmat and regional limits are checked in real time, and shoppers only see the shipping and payment options that apply to them.
Checkout rules that prevent bad orders.
Business rules enforced at checkout, not in the support inbox. Restricted products, PO boxes, hazmat and regional limits are checked in real time, and shoppers only see the shipping and payment options that apply to them.
The problem
Compliance and shipping rules were enforced after the fact, with support cancelling orders that should never have been placed.
What we build
- Cart and checkout validation for restricted products, regions and address types.
- Delivery customizations that hide, rename or reorder rates by cart contents.
- Payment customizations that show the right methods to retail and trade buyers.
Talk to an engineerStack
Pre-order and backorder windows.
Keep selling past zero. Variants switch to pre-order with honest ship dates, orders are held until stock arrives, then release to the warehouse on their own, so demand isn’t lost between purchase orders.
Pre-order and backorder windows.
Keep selling past zero. Variants switch to pre-order with honest ship dates, orders are held until stock arrives, then release to the warehouse on their own, so demand isn’t lost between purchase orders.
The problem
Popular products sell out weeks before the next shipment arrives, and the demand in between is lost.
What we build
- Pre-order and backorder selling by variant, with ship-date messaging on the product page and in the cart.
- Fulfillment holds that release automatically when the purchase order lands.
- Separate deposit or full-payment options, and clear updates to customers when dates move.
Talk to an engineerStack
AI product search.
Search that understands what shoppers mean. Plain-language queries become filters and attributes, results lean on clean product data, and search logs show merchandisers exactly where the catalog falls short.
AI product search.
Search that understands what shoppers mean. Plain-language queries become filters and attributes, results lean on clean product data, and search logs show merchandisers exactly where the catalog falls short.
The problem
Shoppers search the way they talk. Keyword search returns nothing for “waterproof jacket under $300”, and they leave.
What we build
- Natural-language queries turned into structured filters, price ranges and attributes.
- Synonyms and redirects tuned from real search logs.
- Zero-result and low-click reporting, so merchandisers know where to fix the catalog.
Talk to an engineerStack
Flicker-free experiments.
A testing setup built into the theme, so experiments load as fast as the rest of the site. Variants are assigned before render and measured against orders and margin, and every result is logged so wins actually ship.
Flicker-free experiments.
A testing setup built into the theme, so experiments load as fast as the rest of the site. Variants are assigned before render and measured against orders and margin, and every result is logged so wins actually ship.
The problem
Client-side testing tools slow the site and flicker. Results argue about clicks while the real question is revenue.
What we build
- Experiments run on theme sections and templates, assigned before the page renders, with no flicker.
- Results tied to orders and margin, not just clicks.
- A shared log of every test, its result and what shipped.
Talk to an engineerStack
Commerce data pipeline.
Commerce data the whole business can trust. Orders, returns and inventory stream into the warehouse and reconcile with finance, so merch, ops and leadership work from the same numbers.
Commerce data pipeline.
Commerce data the whole business can trust. Orders, returns and inventory stream into the warehouse and reconcile with finance, so merch, ops and leadership work from the same numbers.
The problem
Every team pulls its own report. Shopify, the ERP and the ad platforms disagree, and nobody trusts the totals.
What we build
- An event pipeline for orders, refunds, inventory and customers into the data warehouse.
- Reconciled revenue that matches finance, net of returns, discounts and tax.
- Dashboards for merch, ops and leadership built on the same numbers.
Talk to an engineerStack