Choosing webshop infrastructure: start with what you can afford to lose
At three in the afternoon the checkout starts throwing an error. A plug-in was updated without a test environment, and the quickest fix someone suggests is restoring last night's backup. That brings the checkout back, and it wipes out every order placed today. For a busy webshop that can be hundreds. That afternoon is when you find out what you decided back when you were choosing webshop infrastructure.
So start this choice with the question of what you can afford to lose: how long the shop can be closed and how much recent data can go missing. Only once those two numbers are clear does it make sense to talk about hosting and platforms.
First map what an outage costs you
The cheapest setup is rarely the most economical one. If the shop is down for minutes during a peak, revenue, ad budget and trust all take a hit at once. In a B2B webshop, a slow ordering environment can be enough for customers to fall back on email, the phone or a competitor.
A check for this week: go through each part with a colleague, work out what happens if it drops out for a while, and write the answers down.
- Can a customer still place an order?
- Does the order reach the ERP system correctly?
- Are prices and stock levels still right?
- Who is the first to notice that something is wrong?
Those answers determine how much availability, capacity and oversight you need. A shop with a few dozen orders a day has different requirements from a platform with live stock, product feeds, several countries and thousands of visitors at the same time. Both can run perfectly well on modern technology, each with its own setup.
Take the busiest periods you can see in your own order figures as the starting point. Then you pay for capacity you use and have room on the days the pressure builds.
Choosing webshop infrastructure covers the whole chain
The front end your customer sees relies on a web application, a database, the payment provider, search, images, email and the integrations with stock and accounting. The weakest link often decides what the customer experiences. A fast server does little if an external API makes a slow request on every product page, and a fast integration solves nothing when the database locks up while processing orders. If you only look at the hosting, you miss how it all fits together.
The platform decides how much you manage
Magento, WooCommerce, Shopware, Shopify or custom software: your platform partly decides which infrastructure makes sense. The more technical maintenance the platform takes off your hands, the less room you have for complex processes, custom logic and integrations. A standard SaaS platform takes a lot off your hands. If fast implementation and predictable costs come first, that can be a sensible trade.
With self-hosting or custom software you shape processes around your own operation, build the integrations you need and optimise performance where it matters. It also means updates, security, deployments and monitoring have to be well organised. Freedom without technical ownership quickly becomes a risk.
Our rule of thumb: only choose self-hosting or custom software once you can name who will pick up those tasks. If you can't yet, SaaS is the wiser choice for now.
Measure first, then scale
With predictable traffic, dedicated or managed hosting can be a clear choice. You know the capacity, the environment is easy to oversee and a technical partner can maintain and optimise it in a targeted way.
If your shop has sharp peaks, international visitors or campaigns with a load you can't predict, scalability matters more. Consider separate capacity for the application, the database and the cache, so they don't all lock up at once. Caching and a content delivery network can take load off the main server, especially with large product images and a lot of returning traffic.
We would leave fully automatic scaling aside at first. It can make your costs unpredictable and hide a poorly performing application, so you end up paying for extra capacity to cover a problem in the application itself. Measuring and optimising first, then adding capacity where it is needed, usually pays off more.
Keep heavy jobs away from the checkout
In e-commerce the hardest problems often sit behind the scenes. Stock changes, pricing rules, order statuses, fulfilment, invoices and customer data move between several systems. When that communication falters, the customer may still see a tidy product page while your team ends up fixing orders by hand, at real cost.
You can check this on your current integrations. Does the webshop fetch data live during a visit, or is data synchronised? What happens when the ERP system is temporarily unavailable? And are failed messages processed again, with an alert to someone who acts on it?
Product imports, feed processing and invoice generation belong in background processes, separate from regular web traffic. That keeps the shop available while a lot is happening behind the scenes.
An untested backup is an assumption
Back to that afternoon with the broken checkout. You need a daily backup. For a continuity plan you also need to know how quickly the environment can be restored and how much data loss is acceptable.
So pin down two things. The recovery time: how long can the webshop be down before the damage gets too big? And the recovery point: how much recent data can be missing at most? These are business decisions, and they drive the technical measures that follow, such as frequent backups, replication, separate storage and tested recovery procedures. If you can miss an hour of orders at most, a nightly backup is not enough.
What often goes missing in those decisions is testing. Ask your hosting provider when a restore was last tested. The same goes for updates. Without a test environment, an update can fix a security risk and break your checkout in the same move, exactly as it did that afternoon.
Security starts with who can access what
A webshop processes personal data, order data and often payment information. An environment like that needs active maintenance, with timely security updates, access control, encrypted connections, logging and checks for unusual behaviour.
Give everyone who works on the shop the access that fits what they do. Especially when an agency or several development teams work on the shop, a clear division of roles stops a forgotten account from becoming a way in later. Ask for a list of everyone with admin rights and see which names you can no longer place.
The chain matters here too. An outdated plug-in or an integration nobody looks after can carry more risk than the server itself. Put one party in charge of the overview, even when several suppliers work on the webshop.
Give outages a single owner
With monitoring of performance, error messages, disk space, databases and integrations, you often spot a problem before customers call. If the first alert comes from a customer, you have already lost revenue.
Who acts at that point matters just as much. Developer, hosting provider, platform vendor and integration partner are sometimes separate companies. That becomes a problem when everyone points at each other during an incident, because then the outage lasts longer. Put the responsibilities in writing and choose a point of contact who understands the whole chain.
At LJPc, development and hosting are under one roof. That shortens the path from report to analysis and fix, especially when an error starts where code, configuration and infrastructure meet. If your organisation leans heavily on online sales, that direct responsibility is often worth a lot.
Five questions before you sign
A technical proposal is only useful if it fits how your shop actually runs. Go through these questions up front with every party that makes a proposal:
- What load does the webshop need to handle normally and during peaks?
- Which systems are needed to process an order completely?
- What is the maximum acceptable downtime and the maximum acceptable data loss?
- Who monitors, maintains and restores the environment outside office hours?
- How does the setup grow when the product range, traffic or processes change?
Ask for concrete agreements on capacity, response times, backups, updates, access control and escalation. Then you know what happens when things get tense.
You can start on the third question today. Ask your current hosting provider how often a backup is taken. Then count how many orders come in within one such interval on a busy day. That is how many orders an outage could cost you, and with that number the conversation about your new setup gets a lot more concrete.