Business software

How long does it really take to build business software?

Quentin Debray
Written by Quentin Debray
Published 30 September 2026 · 10 min read
Article updated 21 October 2026

In brief

Three weeks or four months: the gap between two business software projects almost never comes from the number of screens. Here are the seven stages of a schedule, what shortens them, and the number one cause of delay.

When a business owner asks how long it takes to develop business software, the answer they most often get is "it depends". That is true. It is also useless. What you need to know is not the average duration of a project that is not yours, but what the schedule depends on, stage by stage, and which of those levers are in your hands.

This article walks through the seven stages of a project at the builder, what shortens them, what lengthens them, and why the fastest quote is not always the best signal.

Why "it depends" is not an answer

Two projects billed at the same price can take three weeks or four months. The gap almost never comes from the number of screens. It comes from three things:

  • How clear the business process is. A company that can describe its order flow in ten minutes saves two weeks. A company where three people give three different versions first has to decide.
  • The state of the existing data. Importing a clean file is quick. Importing four workbooks with duplicates, free-form columns and varying naming conventions takes a time that is systematically underestimated.
  • The speed of decision on the company's side. This is the number one factor, and we come back to it below because it deserves to be said frankly.

A custom project schedule is therefore not a duration to guess. It is a sequence of stages, each with its own accelerators and its own pitfalls.

The 7 stages that make up the development time of business software

1. Scoping: understanding the process and deciding what we will not do

Scoping serves three purposes: understanding how the work is really done, identifying the bottleneck (the place where your team loses time every week), and arbitrating the perimeter of the first version.

That last point matters most for the schedule. A successful scoping produces as many "no"s as "yes"es. Each function removed from version 1 is not a lost function: it is a function pushed back until after go-live, when you will know whether it is really needed.

What shortens it: a single contact, a decision-maker, who is available. A two-hour meeting with the person who knows the field is worth five email exchanges.

What lengthens it: a scoping run with four departments that do not share the same priorities, with no arbiter. Or a perimeter that grows at every meeting.

2. Design: the screens, the rules, and what happens when things go wrong

We draw the screens, write the business rules, and above all handle the exception cases. This is where schedule overruns hide.

A concrete example: a quote turns into an order. Fine. But what happens if the customer approves half the quote? If the price has changed in the meantime? If the salesperson has left? Each of these questions is a rule to write. Discovered in design, it costs an hour. Discovered in acceptance testing, it costs three days and a schedule slip.

What shortens it: agreeing to spend time on exceptions now, even when it seems fussy.

What lengthens it: "we'll see in use". You will not see in use, you will see at the worst moment.

3. Data: the model, and carrying over what exists

This is the recurring surprise of schedules. The data model (how customers, jobs, parts and interventions are organised) is built fairly quickly when the process is clear. Carrying over the history, on the other hand, is unpredictable.

If your information has lived in hand-kept tables for eight years, it has to be cleaned before it is carried over. Duplicates, varying spellings, columns added by someone who is no longer there, free notes that in fact contain structuring information. This work is feasible, but it is counted in days, sometimes weeks, and depends above all on your availability to settle the ambiguous cases.

What shortens it: a clean export, and someone on the company's side able to say "this line, we keep it" or "we drop it".

What lengthens it: wanting to carry everything over. Often, carrying over the last three years and archiving the rest is enough and saves a month.

If your starting point is not a spreadsheet but an existing application that no longer keeps up, the subject is different and deserves its own diagnosis. We covered it in taking over a business application that no longer keeps up.

4. Development

This is the stage everyone imagines as the longest. It no longer really is. With current building methods, producing the screens and the rules moves much faster than a few years ago, provided the three previous stages are done.

A development that slips is almost never a production problem. It is an incomplete scoping coming back to the surface.

What shortens it: deliveries in batches, shown every week or every two weeks. You see, you correct early.

What lengthens it: a two-month tunnel followed by a big reveal. That is when all the gaps are discovered at once.

5. Testing and acceptance with real users

Acceptance testing is not an end-of-project formality. It is the moment when the people who will use the tool every day try to break it with their real cases.

This stage is almost always under-resourced in schedules. Not by the builder: by the company, which has not set aside time for its teams to test. The result is that acceptance drags over three weeks instead of five days, not because there are many fixes, but because nobody had time to look.

What shortens it: two or three named users, with slots blocked in their calendars, and real cases prepared in advance.

What lengthens it: acceptance entrusted to "everyone", therefore to nobody.

6. Training

It is short when the software follows the company's vocabulary and the sense of the real work. It gets longer when the tool imposes a logic foreign to the team. It is a good revealer: training that drags on often means a design that missed something.

Expect little classroom training, and more support in the first days of use. That is when the real questions arrive.

7. Go-live and the first weeks

Switchover day is not the end of the project. The two to four weeks that follow are a stage in their own right: ergonomic adjustments, forgotten cases, permission settings, small automations nobody saw coming.

An honest schedule reserves this time. A schedule that stops at the go-live date simply pushes the adjustments onto a later invoice, or onto the teams' backs.

The weight of each stage in the schedule

The proportions below are orders of magnitude from the practice of internal projects in SMEs, not a measured figure. They help split up a schedule, not guarantee it.

StageShare of the scheduleWhat shortens itWhat lengthens it
ScopingSmall but decisiveAn available decision-makerA growing perimeter
DesignSignificantHandling exceptions straight away"We'll see in use"
Data and migrationHighly variableClean export, limited history perimeterLate cleaning
DevelopmentThe bulk of the volumeDeliveries in batchesTunnel effect
Testing and acceptanceOften underestimatedNamed users, blocked slotsAcceptance with no lead
TrainingShortBusiness vocabulary respectedImposed logic
Go-live and adjustments2 to 4 weeksTime reserved in the scheduleProject declared finished too early

What it looks like in practice at Magram

Ateliers d'Art (art education): 8 weeks of development for a complete platform, from the public website to the back office including the families area, in other words a complex application. On enrolment day, the tool absorbed 390 registrations, 75% of a month's volume in a single day.

On our last three projects, the build took 5 to 6 weeks of development, delivered in batches. Each batch was shown and approved before moving on to the next.

The first cause of delay is on your side, and that is good news

Let us say it plainly: in most projects that slip, the cause is not technical. It is availability and speed of decision on the company's side.

A screen approval that waits ten days is ten days of project. A scoping meeting postponed twice is three weeks. A choice of invoicing rule that nobody dares settle is a development on hold.

It is good news because this lever belongs to you. Three decisions are enough to secure a schedule:

  1. Name a single lead on the company's side, with the power to decide without escalating every time.
  2. Block the approval and acceptance slots in the calendars in advance.
  3. Accept that a deliberately reduced version 1 goes out quickly, rather than a complete version that may go out.

Why the shortest timeline is not always a good sign

When you compare two proposals and one announces three weeks where the other announces two months, the first is not necessarily the better. Ask three questions.

  • Is scoping included, or billed afterwards? A very short schedule often starts after scoping, which is then invisible in the announced figure.
  • Is data migration counted? It is the most frequent exclusion. And the most costly in time.
  • Are acceptance and post-go-live adjustments within the timeline? If the announced date is that of technical delivery and not of real use, you are not comparing the same thing.

The same reasoning applies to amounts: two quotes that do not include the same perimeter cannot be compared. We detail this mechanism in our analysis of the price of custom software development.

What current methods really change in the schedule

Production has sped up. Screens, rules, sequences: what used to take weeks is now built in days on many subjects. We detailed this in our article on AI-assisted development.

But the speed-up affects only part of the schedule. Understanding a business process, arbitrating a perimeter, cleaning a history, getting busy users to test, training a team: none of that has been cut in half. That is why a serious project does not shrink in duration indefinitely. Human time remains human time.

In other words: the "production" share of the schedule has melted away, the "decision" share has become proportionally heavier. Which reinforces the previous point even more.

How to get a reliable date for your project

A credible estimate requires four pieces of information:

  • The process to be equipped, described from start to finish, with its known exceptions.
  • The state of the existing data, and the history perimeter you really want to carry over.
  • The number of users and their roles, because access rights are part of the work.
  • Non-negotiable external deadlines (year-end close, peak season, audit, a site move).

With these four elements, a schedule is laid out in dated stages, with the approval points you are responsible for. Without them, any announced date is a polite illusion.

Where to start

If you are looking for a duration before launching a project, start with the diagnosis and not the quote. An hour's conversation about your process is generally enough to say whether the subject can be handled in a few weeks or needs a broader scoping, and to identify straight away the two or three points that will tip the schedule.

That is exactly the purpose of a first conversation with the builder: looking at the bottleneck, setting the minimum useful perimeter, and leaving with dated stages rather than an "it depends".

Also read: What is custom business software?

Also read: Build or buy software: how to decide

A question about your situation?

Ask Quentin, he replies personally.

Your details are used only to reply to you. Learn more

We use audience-measurement cookies to improve the site. Learn more.