Custom Ecommerce Software, from Storefront to Back Office

Storefront work on Shopify

The storefront is where the platform does most of the work, and where custom work should be targeted. We build custom storefront features on Shopify when a theme setting or an app will not do: product pages that handle configurable or bundled products, account areas that show order and subscription history, and custom apps that connect the store to the rest of your stack. Payments run through the platform's checkout or Stripe, depending on the build.

The back office: where custom software fits behind the store

Behind the store, the work worth automating is almost always the work between systems: the steps where someone copies a fact from one screen to another, or reconciles two numbers that should have matched.

• →Order routing and sync. Orders from the store and each marketplace land in one place, pass your fraud and address rules, and reach the ERP and the warehouse or 3PL without anyone rekeying them.
• →Inventory that agrees across channels. One record of available stock, pushed to every channel on a schedule you set, with buffers per channel so a marketplace does not oversell the last unit.
• →Catalog data cleanup. Product titles, attributes and descriptions normalized to one standard, with a language model proposing the fix and a merchandiser approving it before it publishes.
• →Returns and exchanges. A return request checked against your policy, the right label issued, and the refund, restock or write-off recorded in every system it touches.
• →Support ticket triage. "Where is my order" and "how do I return this" answered from the order record, and everything else routed to your support team with the order history already attached.

The system of record problem

Every store that sells in more than one place eventually has the same argument: which system is right? The storefront says ten in stock, the warehouse says eight, the marketplace listing says twelve, and the ERP has not been updated since Tuesday.

The fix is not another sync app bolted onto each pair of systems. It is deciding, per piece of data, which system owns it: the ERP owns cost, the warehouse owns on-hand quantity, the platform owns the customer's cart. Every other system reports its changes to the owner and never overwrites it: a channel posts the sale, the warehouse adjusts on-hand, and each channel's available count follows from the owner. Those decisions have to be settled before the build, because every integration that follows depends on them. It is the easiest step to skip and the one that decides whether the rest works.

Where AI helps, and where it does not

AI is useful in ecommerce operations in a few specific places. It is good at turning a supplier's inconsistent product spreadsheet into clean, consistent attributes. It is good at reading a support email and working out which order and which question it is about. It is good at forecasting demand when there is enough order history to learn from.

It is not a substitute for clean data or clear rules. A model will not tell you which system owns inventory, and it should not decide a refund on its own. So the AI steps we build can propose rather than act, depending on the workflow: the merchandiser approves the catalog change and the support agent approves the reply. Where the engagement calls for it, the output is also checked in code against your records and rules before it reaches a reviewer: an attribute outside your allowed values, a reply that cites an order the customer does not have, or a refund that does not match the order is flagged, not queued as a normal proposal. The model's own sense of how sure it is does not decide anything.

Build or buy: most stores should start with the app store

Shopify, the other major ecommerce platforms and the major marketplaces have large ecosystems of apps for reviews, subscriptions, shipping labels, simple inventory sync and more. If an app does what you need, use it. It will be cheaper, faster and maintained by someone whose whole business is that one feature.

Custom software makes sense when the workflow crosses systems the apps do not connect, when your rules are specific enough that every app needs a workaround, or when you are paying for a stack of apps that each solve a third of the problem. Our build vs. buy guide covers the decision in more detail.

Retail clients

Candlefish Library, the retail and class-booking experience for a Charleston candle brand's library of 100 fragrances, is one of the clients we name publicly.

How an ecommerce engagement runs

We start with the 90-minute AI workflow audit: one session tracing an order from checkout to delivery to return, and marking every place a person touches it. The audit is complimentary when it is bundled with a larger build and available as a paid one-off on its own.

From that we write a priced scope for the one flow that costs the most hands, such as order sync or returns. Prototype, staging, and deployment milestones are agreed in the written scope. Before the build, we agree what better means (orders keyed by hand per week, or oversells per month) and you record today's number, so the result can be checked against it.