Start a project

Why software testing matters more now than it used to

4 min read

Software ships faster, runs on more devices and is increasingly written with AI. Here is why testing has become more important, and what skipping it costs.

A software team working late around laptops in a glass-walled office

Ten years ago a company might release its software a few times a year. Each release was planned for months, and there was time to check it. That world has gone. Here is what changed, and why it makes testing matter more, not less.

1. Software ships far more often

Many teams now release every week, and some release several times a day. That is good for customers, who get fixes and features quickly. It also means there is no longer a quiet month before launch in which someone checks everything.

When you release fifty times a year, you have fifty chances to break something that worked last week. A check that is not built into the release process simply does not happen.

2. There are more ways to use your product than ever

Your customer might be on a five-year-old Android phone, a new iPhone, a tablet in landscape, or a laptop with the browser zoomed to 150%. They might be on fast fibre or on one bar of mobile signal on a train.

A feature that works perfectly on the developer's own laptop has been tested in exactly one of those situations. The others are where real customers find the problems, usually without telling you. They just leave.

3. A growing share of code is written with AI

AI coding tools can produce a working feature in minutes. That speed is real. So is the catch: AI-written code looks confident whether or not it is correct. It handles the example it was shown and can quietly mishandle the case nobody mentioned.

When code is produced faster than a person can read it carefully, checking the result becomes the main safeguard. Teams that write more code with AI need more testing, not less.

4. Your product depends on things you do not control

A modern app is rarely one piece of software. It calls a payment provider, a maps service, an email service, a login provider and a dozen libraries written by other people. Any of them can change.

When a payment provider updates its page, or a browser releases a new version, your product can break without anyone on your team touching it. Only a test that runs regularly will notice.

5. Users have less patience and more choice

If a page takes too long or a button does nothing, switching to a competitor takes one tap. People rarely report a bug. They assume the product is poor and move on, and some of them write a review on the way out.

The first impression is often the only one. A sign-up form that fails on a popular phone does not get a second chance from the people it failed.

6. Mistakes are more public and more costly

A bug used to be an inconvenience. Today software handles payments, medical appointments, travel bookings and personal data. A fault can charge a customer twice, expose private information, or stop a business taking orders during its busiest hour.

Screenshots of failures spread quickly, and in many places a data leak carries legal consequences as well.

What skipping testing really costs

A problem gets more expensive the later it is found:

  • While the feature is being built, fixing it is a few minutes of one developer's time.
  • Before release, it is a bug report and a fix: an hour or two.
  • After release, it is lost sales while it is broken, support messages, an urgent fix, a rushed second release, and the customers who never came back.

The bug is the same in all three cases. Only the bill changes. Testing does not remove the cost of mistakes. It moves the discovery to the point where it is cheapest.

What good testing looks like now

It does not mean a large team or months of work. For most products it means three habits:

  1. A person tests each new feature before release, on the devices and browsers your customers actually use.
  2. The journeys that earn money are checked automatically on every change: sign-up, login, search, checkout, payment.
  3. Every bug that reaches customers becomes a test, so it cannot quietly return.

That is enough to catch most of what would otherwise reach your users.

Where to begin

If you have no testing at all, start with one careful pass over your main journeys. It shows you how much is wrong and which parts deserve automated checks.

We do that first pass as website testing and mobile app testing, and keep the important journeys covered with Playwright automation. To decide which checks belong to people and which to a machine, read manual testing vs automated testing.

Chat with us