Spike - Our Thinking

How should retailers test end-to-end journeys across ecommerce, ERP, warehouse and POS?

Written by Raisa Lupascu | 18 November 2026

If you’re a retailer, test end-to-end journeys by following real orders from the website and store tills (POS) through ERP, warehouse, payment, finance and returns, in a test environment that mirrors production. Start with one simple order that travels the whole way, then add the variations: delivery types, payment methods, cancellations and returns. Prioritise the journeys that bring in revenue or would cost the most if they failed, and start as soon as your systems begin to line up.

Every system can pass its own tests and the journey can still fail. The gaps between systems are where retailers get caught out, so that’s where testing needs to look.

A customer clicking “buy” looks like one action. Behind it, hundreds of processes run in the background. Stock is held, payment is taken, the warehouse picks and ships, emails go out, and the sale lands in the ledger. If the customer sends it back, the whole thing runs again in reverse.

Each of those steps can sit in a different system, often built or run by a different supplier. That’s where the risk sits.

When we tested Lovehoney’s move to NetSuite, those journeys ran across more than 50 integration points, from Salesforce Commerce Cloud through NetSuite, order management, warehousing and finance. We automated them as real customer journeys, from storefront to fulfilment and finance.

We use NetSuite as the worked example here, because it so often sits in the middle of these journeys. The approach applies to any retail stack. If NetSuite is your ERP, our guide to NetSuite testing covers implementations and upgrades in more detail.

Why journeys fail when every system has passed its own tests

Most failures happen in the handoffs, where one system passes something to the next. In our experience, four causes come up again and again.

1. Integrations and data

An integration doesn’t work as designed, or one system expects data from another in a set format and it isn’t there. A field mapping is missed, or a value arrives in the wrong shape. Each system looks fine on its own.

2. Reference data

Products need all their mandatory data and feeds set up before an order can flow. If a product is incomplete in one system, every journey that touches it can fail.

3. Missing requirements

If a business scenario was never written down, it was never tested. The gap only shows up when a real customer does it in production.

This is where experience counts. An experienced tester reads the requirements and asks the questions nobody has asked yet. What happens with a return? What if the customer cancels, or the retailer does? It’s one reason we bring QA in early, as we explain in Why testing before coding is the smartest move in retail tech.

4. Test environments that don’t match production

Not everything is fully simulated in test. A team may enter data partway through a journey, while in production that data comes from a system further upstream. The test passes, and the live journey fails at a step nobody tested.

Edge cases add to all four. These are the orders that aren’t the norm: a basket with 60 products in it, or home delivery and store pickup in the same order. Most customers never do them, but some will.

We see the same pattern on Shopify projects, where most of the risk sits in the integration layer.

Which journeys to test end to end

The core set follows an order through its whole life, from the customer’s account to the ledger, and back again for returns. These are the journeys we include for almost every retailer.

Journey What to check across systems
Customer registration New customers can sign up and log in, and their record reaches every downstream system
Order placement Different product types, delivery types and payment methods
In-store sales (POS) Sales and payments taken at the till reach stock, finance and the customer record, the same as online orders
Stock and availability Adding to basket holds stock, the last item can only be sold once, and the website matches what the warehouse and stores hold
Payment Every payment method you offer, online and in store, is processed correctly in each system after the order is placed
Fulfilment and delivery The order reaches the warehouse, 3PL or store for click and collect, and tracking and customer emails go out at each stage
Finance Every order lands in the general ledger and reconciles with the warehouse, stores and shipping
Cancellations and returns Whoever cancels, and wherever an item comes back, including online orders returned in store, the money goes back and the stock goes back on sale

 

Start with the simplest version of each journey to prove the basic integrations work. Then add the variations. A fridge and an iPhone don’t travel the same way: a customer can pick up a phone in store and take it home, but a fridge needs a delivery.

What often gets left out

Payment methods are the most common gap. Teams test card payments heavily, but PayPal and marketplace payments often get less attention, usually because the test environment doesn’t have those integrations. The order may place fine and then fail in the next system, because not every system is set up for every payment type.

Product data is the other. Most testing assumes products already exist, because creating them involves feeds from several systems plus enrichment such as photography. That’s reasonable if your products arrive from suppliers’ systems. If your own teams create and maintain product data, test that too.

 

What needs to be in place first

You need four things before you can test end to end:

  1. Requirements that cover every scenario a customer or colleague could trigger, including the awkward ones.
  2. An environment where all the systems in the journey are connected and set up like production. Where an integration isn’t available in test, record it as a known risk.
  3. Test data that can flow through every system. A new process usually needs new data created in every system. A change to an existing process can often reuse existing test data.
  4. Code that’s ready, with each supplier’s work delivered into the same environment.

Automation helps with the data. A good automated test creates its own customer at the start, so it doesn’t depend on a customer already existing in every environment. Automation also makes regular end-to-end runs affordable. Our NetSuite regression calculator shows how long a full cycle takes by hand compared with automated.

These four rarely all arrive on time. When something is late, work on what’s available, and be clear about what’s blocked and why. That keeps testing moving without hiding the gaps.

How to prioritise when you can’t test everything

Agree the priorities with the business, and start with the journeys that bring in the most revenue or would cause the biggest loss if they broke. For most retailers that’s placing an order, taking payment and getting it delivered. We call this the golden thread.

Lower down the list are journeys a contact centre could handle by hand in an emergency, such as customers updating their addresses, saved cards or preferences. You’d never want these to fail, but customers can still buy and receive their orders if they do.

If coverage has to be cut, make the decision visible. List what won’t be tested, set out the risks, and get the business to agree them. We’ve seen retailers go live with only the vital journeys tested and fix forward from there. Done deliberately, that can be a strong decision, because it keeps everyone focused on what has to work on day one.

For more on making that call, see When is “good enough” good enough? and our Graded Test Approach: test in proportion to what happens if it goes wrong.

When end-to-end testing should run

It depends on the size of the change and how mature your IT is.

In a big transformation, such as a new ERP or ecommerce platform, start end-to-end testing as soon as several systems begin to line up. Get one simple order from start to finish to prove the basic integrations are there, then build out the variations. We cover this kind of programme in more detail on our Land Big Change page.

In a mature environment making smaller changes, decide case by case. Look at what the change is, what it touches and the risk if it goes wrong. A well-understood change to one system may not need every journey run again. Every release should still get at least a smoke test of the core journeys, because you can’t always be sure what else went into it.

How your teams release matters too. If several teams push changes into the same release, run the journeys end to end. If it’s a tightly controlled change and you know nothing else went in with it, a lighter check may be enough. That only works where development, QA and operations communicate well.

Platform upgrades are a common trigger. Our article on what needs testing when NetSuite releases an upgrade covers this in detail.

Where NetSuite fits in

It depends on which NetSuite modules you’ve licensed and which systems you already run. Some retailers use NetSuite for almost everything. Others keep some systems and move only part of the business onto it.

In the retailers we’ve worked with, a few patterns come up:

  • Finance almost always sits in NetSuite, including accounting and the general ledger. So nearly every journey ends there.
  • Customer, stock and order data often live in NetSuite too.
  • Warehouse management is often separate. Retailers tend to keep an existing warehouse system, or use third-party logistics partners, so NetSuite exchanges information with them.
  • The ecommerce platform, such as Shopify or Magento, sends orders in and receives stock and status updates back.
  • Store tills (POS) send sales, returns and stock movements in, so stores, website and finance work from the same numbers.

That makes every NetSuite integration a handoff worth testing, whether you’re moving to NetSuite or already running on it. Before you plan your journeys, map which system owns each step for your business, and where the data passes between them. You can’t assume it from another retailer’s set-up.

Who needs to be involved

Most retail stacks are built and run by several suppliers and internal teams, and every journey crosses at least one handoff between them. A change to the website can affect the ERP. So before any change, work out your dependencies: which teams and suppliers are involved, who is responsible for what, and what the business expects to gain.

Then bring QA in early, alongside UAT and operational support. There’s landing a change and supporting it, and one doesn’t work without the other. A two-week testing window at the end of the project plan leaves no time to find the scenarios nobody wrote down. For the UAT side, see Why UAT hurts and what good actually looks like.

Where Spike fits in

We help retailers test the journeys that run across their systems, from the first order on a new platform to every release after go-live, including independent NetSuite testing.

For Lovehoney, we maintain an automated regression pack that runs from the storefront order through to the general ledger. We also ran independent end-to-end testing for Ooni’s move to NetSuite, covering orders from Shopify through to their 3PL partners.

If you’re planning a transformation and aren’t sure which journeys to test first, now is the time to work it out, while there’s still room in the plan.

Read the Lovehoney implementation case study

Frequently asked questions