Business software

Build or buy software: how to decide

Marine Debray
Written by Marine Debray
Published 30 August 2026 · 11 min read
Article updated 25 September 2026

In brief

A market subscription or an application built for your business: the real question is not the entry price but the cost over five years and how specific your process is. A decision grid in three questions, the cases where buying clearly wins, and the third option nobody asks about.

Build or buy software: the question almost always arrives at the same moment. A spreadsheet has grown, a team has doubled, a customer asks for tracking you cannot produce. You look at the market, you see thirty vendors who all promise to solve your problem, and you wonder whether yours is really the one they describe. This article does not defend one route against the other. It gives you a grid for deciding, with the cases where buying is clearly the right choice, those where building is, and a third option almost nobody asks about.

Let us say it straight away: we build custom applications. And we regularly tell business owners to buy market software. A builder who always answers "build" is not an adviser, it is a salesperson.

Placing yourself with three questions

Before comparing prices, answer these three questions. In the vast majority of cases they are enough to steer the decision.

1. Is your need standard, or specific to the way you work?

Standard means: any company of your size has exactly the same need, expressed in the same way. Payroll. Accounting. Electronic signature. E-invoicing. These areas are standardised, often by law, and you have no interest in reinventing a rule that the State rewrites every two years.

Specific means: your way of doing things is what sets you apart. Your pricing method, your planning logic, your way of following a customer file from end to end. There, a generic tool asks you to give up what works.

2. How many people will use it, and for how long?

Five users on a secondary need is not the same calculation as thirty people who live in it eight hours a day. The more users there are, the heavier the subscription weighs and the more rational custom becomes. We come back to this with figures below.

3. Do you keep a file ALONGSIDE your current software?

This is the most reliable sign there is. If your teams maintain a spreadsheet in parallel to make up for what the tool cannot do, the question is already settled. You pay a subscription, and you also pay human time to plug the holes. That spreadsheet is the specification of your future application, written for free by the people who know the business.

The cases where buying wins, no discussion

We have to say this frankly, otherwise nothing that follows is credible.

The need is standard and regulated. Payroll, accounting, social declarations, e-invoicing. The regulations change, and market vendors absorb that change for you. Building in there means committing to follow the law yourself. Nobody wants that contract.

The market is mature, with several serious vendors. When five established players have been fighting over the same segment for ten years, the functions have aligned, prices have settled and documentation exists. You are buying a product run in by thousands of users before you.

You have fewer than ten users. At that volume, a subscription remains bearable and custom struggles to pay off, unless the need is really very particular.

Nothing in the process sets you apart. If you do as everyone does and it suits you, take what everyone takes. Custom has a cost in attention: you have to decide, arbitrate, test. Do not spend that attention on a subject that is not what makes your value.

A recent example: an electrician came to see us with a tool project modelled on Qonto. We advised him to take Qonto directly. What he wanted already existed.

The cases where building wins

You bend your way of working to fit the tool. You changed a process that worked, not because it was bad, but because the software could not handle it. The cost of that bending is invisible in the budget and very real in your teams' days.

You maintain compensation files. The point seen above. A tracking spreadsheet, a second export reworked by hand, a shared document that holds the real information. Every compensation is a confession.

Your process is your competitive advantage. You reply faster, you price better, you follow more finely. Putting that in a standard tool means lining up with your competitors who use the same one.

Your data is scattered across five tools that do not talk to each other. People spend their time copying. You are no longer paying for software, you are paying for re-keying.

The need combines two or three trades. No vendor covers the intersection, so you buy three subscriptions and make them hold together with goodwill.

If you want to understand what an application built for a specific business looks like in practice, the article what is custom business software details what it covers and what it does not.

The opposite happens too. A company we are talking to at the moment only half uses Salesforce. Every extension is paid for on top, the team exceeds 150 licences, and a price rise is coming that it does not want to absorb. It is studying a custom CRM, designed for its teams and integrating AI. It has not decided yet.

Another case, this one decided. A property management company was working with Ines, a market software whose licence is coming to an end. It first looked at HubSpot, then ruled it out for two reasons. The cost per licence climbs with the size of the team, whereas a custom tool is paid for at a fixed price. And HubSpot is designed for sales, not for what follows: after-sales and sales administration. It chose to build its tool with us.

The third option: buy, then connect

The question "build or buy software" is badly put when it assumes you have to choose for the whole company. In practice, the most economical answer is often mixed.

You keep the market software where it is unbeatable: accounting, payroll, invoicing, signature. And you build only the missing piece, the one that carries your specificity, making it communicate with what exists.

An example of a typical split in an SME:

AreaRecommended routeWhy
Accounting, payrollMarket softwareRegulated, legal updates included
Invoicing, signatureMarket softwareStandard, strong legal constraints
Job tracking, pricing, workshop planningCustomCarries your method, no tool fits
Customer portal, file trackingCustomDifferentiating, closely tied to your organisation
Email, document storageMarket softwareCommodity, no point in building

This split greatly reduces the perimeter to build. You are not financing an overhaul of your whole system: you are financing the half that exists nowhere. And you stop paying for three subscriptions to approach a need that a single tool, cut for you, covers entirely.

Comparing costs over five years, not the entry price

This is where most trade-offs go wrong. A monthly subscription is compared to an investment, which makes no sense. You have to set the total cost of ownership over the real lifespan of the tool, five years at least.

Take a subscription at €45 per user per month, a common price for a management tool in 2026, for a team of thirty.

HorizonSubscription cost (30 users)
1 month€1,350
1 year€16,200
3 years€48,600
5 years€81,000

€81,000 over five years, and you own nothing at the end. Opposite, a custom business application of reasonable perimeter commonly falls in a range of €5,000 to €25,000, paid once, plus a hosting and maintenance cost. The detail of what makes that range vary is in our article on the price of custom software development.

Careful, this table does not say "always build". It says three things.

  1. Below ten users, the subscription is often still the winning calculation.
  2. Beyond twenty or thirty users, the gap becomes such that the question deserves to be asked seriously.
  3. What weighs heaviest appears in neither of these two figures: the time your teams spend compensating for a poorly fitted tool. One hour per person per day across thirty people is a cost item that far exceeds both columns.

The items forgotten on both sides

On the buying side: initial configuration fees, training, add-on modules billed separately, price rises at each renewal, the exit cost if you want to get your data back.

On the custom side: hosting, maintenance, the changes you will inevitably ask for, and your teams' time during the scoping phase. That time is not wasted, it is what makes the tool match the business, but it has to be budgeted.

What you lose by building

An honest article has to set out the trade-offs.

No community. Nobody has written a tutorial on your application. No forum, no video, no third-party consultant who already knows it. The documentation is what the builder produces.

No automatic regulatory update. If a text changes and your application is affected, someone has to step in. That is exactly why we avoid building on payroll or accounting.

Dependence on the builder if the code is not handed over. This is the point to lock down by contract. Always ask: who owns the code, in what form it is handed over, and who can take over the work if you change provider. An application for which you have neither the code nor the access locks you in as surely as a vendor.

No trial before paying. A subscription can be tried for two weeks. An application is scoped then built. Hence the importance of starting small, on a reduced perimeter, rather than launching an eighteen-month project.

Conversely, vendor dependence exists on the buying side too, and it is more insidious. A unilateral price rise, a function removed from a plan, a roadmap heading in a direction that does not concern you, the vendor being bought out. You have no leverage.

Three frequent mistakes in this trade-off

Writing a hundred-page specification before deciding. The document becomes a catalogue of wishes, no tool meets it 100%, and you choose the one that ticks the most boxes rather than the one that really solves the problem. Start with the question: what costs us the most today?

Confusing generic management software and a tool for your business. An off-the-shelf package covers a broad area well, in an average way. If you are trying to understand where the boundary lies, the gap between a package and dedicated development is mostly about how closely it fits the process.

Believing custom has become free. AI-assisted development methods have genuinely shortened timelines and lowered the entry ticket. What has not changed: the scoping time, the understanding of the business, the trade-offs. We detail this shift in what changes and what does not when you develop with AI.

How to decide this week

A simple method, applicable in a few days.

  1. List the parallel files your teams maintain alongside the tools in place. Each file is a symptom, note who keeps it and how long it takes.
  2. Count the real number of users per tool, and calculate the cumulative bill over five years.
  3. Split your needs into two columns: regulated/standard on one side, specific to your method on the other.
  4. On the standard column, go and look at the market. On the specific column, test two or three tools, and if you have to bend your process to fit them, stop the test.
  5. Put a figure on the smallest possible perimeter on the custom side. Not the global overhaul: the piece that unblocks the rest.

Return on investment is rarely measured on the licence saved. It is measured on the re-keying hours removed, the errors avoided, the shortened customer response times. Those are the figures to put on the table before choosing a route.

If after this exercise you are still hesitating, it is often because the answer is mixed: buy what is standard, build what sets you apart. And if market software does the job, take it. We will tell you so.

Taking stock with Magram

The three questions at the start (is your need standard or specific, how many people will use it and for how long, do you keep a file alongside your current software) are settled faster in two than in an internal meeting. That is what the Magram diagnosis does: about ten minutes, a costed order of magnitude, and your needs sorted into two columns, the one you buy and the one you build.

You leave with a frank opinion, including when that opinion is "market software will be enough". It is the one in this article, applied to your organisation rather than in general.

Take the diagnosis

Let's talk about your projectQuick replyNo obligation

Also read: Off-the-shelf package vs software: differences, uses, choices and trade-offs in 2026

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.