Ecommerce Tech Stack: What to Buy First and What to Defer
Most ecommerce tech stack advice is a list of logos. That is useless, because the tools are interchangeable and the decisions are not. A stack is five layers, one of which you cannot change without a rebuild, and four of which you can swap in a weekend. Get the sequence right and the tooling budget takes care of itself. Get it wrong and you replatform in year two.
Don't pay full price to build this.All 39 Ecommerce tools below come with a discount code, a startup programme or free credits.Claim these perks →
The five layers, and the one you cannot change later
An ecommerce stack is five layers, bought in a predictable order: the store platform that renders the catalogue and takes the order, the money layer that captures and settles it, the retention layer that brings the customer back, the fulfilment layer that gets the box moving, and the back office that reconciles all of it. Every tool you will ever evaluate is a feature of one of those five.
Only one layer is genuinely load-bearing. Swapping an email provider is a weekend of list migration and a resent welcome flow. Swapping the store platform is a front-end rebuild, a new URL map, a redirect table, a re-integration of every app you bolted on, and a fortnight of ranking volatility while Google recrawls. Choose the platform on a three-year view and treat the other four as provisional.
That asymmetry sets the sequence. Spend real time on the platform and the payment provider behind it, then buy the rest cheaply and reversibly. Founders routinely invert this, agonising over which review widget to install while accepting a platform that cannot model their product variants. The widget swaps in an afternoon. The variant model does not.
Picking a store platform you will still be on in three years
The hosted versus self-hosted argument is really an argument about who absorbs operational cost. Hosted platforms take PCI scope, uptime, patching and checkout maintenance off your plate and charge a percentage for it. Open-source platforms hand you the code and the bill for a developer on retainer. Below roughly one product manager's worth of complexity, hosted wins on total cost every time, and the percentage is cheaper than the salary.
Lock-in is not the subscription, it is the front end. Themes accumulate app-injected snippets, custom metafields, hand-edited templates and merchandising logic that exists nowhere else. None of that is portable. Migration means rebuilding the storefront, remapping product and collection URLs, and shipping a redirect table before launch, not after. Teams that skip the redirect table lose their category rankings and spend six months earning them back.
The variable that should actually decide the platform is catalogue complexity. Simple products with a size and a colour run anywhere. Configurable bundles, made-to-order options, per-variant weights and dimensions, subscriptions, or B2B price lists narrow the field fast, and they are exactly the things a plugin papers over badly. Model your hardest product on a trial store before you commit, not your easiest.
Payments, and the surcharge nobody models
Hosted platforms charge a transaction surcharge when you process through a third-party gateway rather than their native one. It is small per order and structural at volume, and it is the single line item most founders never model. Before you pick a gateway on features, work out the surcharge on your projected annual gross merchandise value, because that number frequently exceeds the entire rest of the software budget.
The second cost is chargebacks, and it scales with paid social. Cold impulse traffic charges back at a materially higher rate than search or email traffic, and processors watch your ratio, not your absolute count. Cross a threshold and you get reserves, delayed payouts, or termination. The defences are dull and effective: clear billing descriptors, delivery estimates on the product page, tracking numbers pushed to the customer, and a refund policy that is easier to use than a dispute.
Then there is the money underneath the gateway. Payouts arrive net of fees on a rolling schedule, in the currency of the transaction, and your supplier invoices are due in another one. Keep the trading account separate from the personal one from day one, and pick a banking layer that handles multi-currency without a conversion on every leg. This is the layer where a bad choice quietly costs a margin point on every international order.
Retention, or why the acquisition maths works at all
Paid acquisition rarely pays back on the first order. It pays back on the second and third, which makes email and SMS the only channel that makes the rest of the stack viable. Install the retention tool before launch, not after, because the list you fail to collect during your first traffic spike is gone permanently.
The mechanism that catches people out is the sending domain. A brand-new domain has no sending reputation, so a first campaign blasted to a bought or long-dormant list lands in spam and poisons the domain for months. Authenticate properly with SPF, DKIM and DMARC, send from a subdomain, and warm up with small engaged segments over a couple of weeks. Transactional mail and marketing mail should not share a sending identity.
Reviews belong to the same layer, and the trigger matters more than the tool. Request the review on delivery, not on order confirmation, and ask for a photo, because photo reviews do double duty as conversion proof and as ad creative you did not have to shoot. Reviews also feed product structured data, which is how you get rating stars in search results rather than a plain blue link.
Fulfilment, inventory and the failures that cost real money
Shipping software exists to solve two problems: buying labels at rates better than retail, and never typing an address twice. Run address validation at checkout rather than at label purchase. An unvalidated address is not caught until the carrier returns the parcel, at which point you have paid outbound freight, return freight and a second outbound, and lost a customer who thinks you never shipped.
Inventory goes wrong the moment you add a second sales channel. Two systems both believing they own the stock count will oversell during any spike, and the resulting cancellations are penalised much more harshly on marketplaces than a listing that was simply out of stock. Decide early which system is the source of truth for quantity, push in one direction only, and buffer the exposed channel rather than syncing to zero.
The back office is the layer everyone defers until it becomes an emergency. Gateway payouts arrive batched and net of fees, so they never match your order totals, and reconciling them by hand stops being possible at a few hundred orders a month. Sales-tax registration thresholds also arrive before you notice, and they are triggered by where your customers are, not where you are. Connect accounting to the store early and let it accrue the history.
Sequencing: week one, month three, and later
Week one is deliberately short: a store platform, a payment provider, one shipping account, a business bank account, and a retention tool collecting addresses. That is the whole stack you need to take a first order. Anything else added before you have customers is a subscription bought to feel productive, and it will still be billing you when you pivot.
Month three is where you add what your actual order flow demands: reviews once you have delivered enough orders to ask, a helpdesk once email volume outgrows an inbox, accounting once reconciliation stops fitting in a spreadsheet, and paid channels once you know the margin on a delivered order rather than a placed one. Let the constraint pull the tool in.
Defer subscriptions, loyalty schemes, ERP, headless front ends and custom checkouts until something specific breaks without them. Each adds an integration surface that has to survive every future platform change. The ecommerce section of this directory lists the founder programmes, partner rates and startup offers behind the tools in each of these five layers, which is worth checking before you commit, because first-year pricing on a stack varies enormously depending on which door you walk in through.
The offers mentioned above
Pulled live from the directory, so what you see here is what the listing says.
Mercury
Finance & Accounting
$250 sign-up bonus
OtherLink on plan
Also mentioned on this page with nothing currently on offer: Shopify, Stripe, Klaviyo, Loox, ShipStation, QuickBooks (Intuit). We track them so you do not have to go looking.
Ecommerce codes, all in one place
39 listings in Ecommerce, with the redemption codes and programme links behind one subscription. Browsing is free.
Common questions
What is the minimum ecommerce tech stack to launch?
A store platform, a payment provider, a shipping account, a business bank account and an email tool collecting addresses. Everything else can wait until an order flow demands it.
Should I use Shopify or an open-source platform?
Hosted platforms trade a percentage for taking PCI scope, uptime and checkout maintenance off you. Open source is worth it once you have a developer on retainer or a catalogue that hosted apps model badly.
How much should an early ecommerce stack cost per month?
Less than you expect if you buy in sequence. Most layers have free tiers or founder programmes at launch volume, and the largest hidden cost is usually the payment surcharge, not the subscriptions.
When should I replatform?
When the catalogue model or checkout is the constraint on revenue, not when the admin annoys you. Replatforming costs a front-end rebuild, a redirect map and a period of ranking volatility.