Information security for software delivery

ISO/IEC 27001
ISO/IEC 27001 certified
Three options
Including the one most comparisons leave out
The third column is where most organisations end up and it is rarely presented as a deliberate choice, which is a shame because choosing it deliberately produces a much better result than arriving at it by accident.
| Option | What you are buying | Where the cost really sits | What you give up | The failure mode | Best for |
|---|---|---|---|---|---|
| Configure a package | Somebody else's model of how the process works, plus their maintenance and their roadmap. | Configuration, integration and data migration. The licence is usually the smaller half over five years. | The processes that differ from the package's assumptions. Some of those are worth keeping. | Customisation to close the gap, until the package is bespoke software you cannot upgrade. | Processes where being standard is fine — finance, HR, ticketing, most back-office work. |
| Build bespoke | A system that matches your process exactly, and an ongoing obligation to maintain it. | Build, then maintenance for as long as it runs. The second number is the one people underestimate. | The vendor's roadmap, their security updates, and the option of walking away from it. | Building the differentiating twenty percent and the commodity eighty percent at the same price. | The process that is genuinely how you compete, and regulated workflows no package models. |
| Package plus bespoke edge | A package for the commodity, and a built layer for the part that differentiates. | The licence, plus a smaller build, plus the integration between them — which is a real line item. | Simplicity. Two systems means an interface, and interfaces need owners. | The boundary drifting until the bespoke layer quietly reimplements the package. | Most organisations, most of the time — chosen deliberately rather than arrived at. |
The real question
Which of your processes are actually load-bearing
Every organisation believes some of its processes are distinctive. Most of the time a smaller number are than people think, and the difference between those two sets is what decides this question. A process is load-bearing if changing it would change what customers experience or what the business can charge for. A process that is merely familiar — approvals routed in a particular order, a report laid out a particular way — is a habit, and paying to preserve a habit in software is how a package implementation becomes a bespoke build at package prices.
The test that works is asking what would happen if the process changed to match the package. Where the answer is that a customer would notice, or a regulator would, or a margin would move, keep the process and build. Where the answer is that people would need retraining and would complain for a quarter, the package assumption is probably fine and possibly better than what you have.
This is also why the honest answer is frequently the package. A supplier that builds software has an obvious interest in the other answer, which is worth naming rather than working around. Where a package covers the requirement, saying so costs us a project and saves the client the larger expense of owning code they did not need to own.
What goes wrong
Four ways this decision goes badly
- 01
Customising a package past the point of return
Each change is small and justified, and after two years the package cannot be upgraded without redoing all of them. The organisation now maintains bespoke software and pays a licence for the privilege.
- 02
Comparing licence cost to build cost
The comparison that makes the package look free. Configuration, integration, migration and the internal time to run the implementation routinely exceed the licence over five years, and they belong in the same column.
- 03
Building the commodity part
Authentication, notifications, reporting, user administration. Building these consumes the budget that the differentiating feature needed, and none of them will be as good as the thing you did not buy.
- 04
No exit written into either
With a package: can you get your data out in a usable form, and what does that cost. With bespoke: is the code documented and deployable by someone other than the original team. Both questions are cheap at contract time and expensive at year four.
Our position
We build software, which is why this page says buy
What we do
- Analyse which processes are load-bearing before recommending anything, because that is the input the decision turns on.
- Say when a package covers the requirement, including when that means we do not get the project.
- Build the differentiating layer and the integration where the hybrid is the right answer.
Where the system will hold personal data, residency and processing obligations come from the data protection regulator and apply to a package vendor's hosting as much as to your own. That question narrows the package shortlist and is worth asking before evaluation rather than after.
Primary sourceCommon questions
What buyers ask
How do we cost the two options comparably?+
Put everything in both columns over five years. For the package: licence, configuration, integration, data migration, internal implementation time, and the renewal escalation in the contract. For bespoke: build, then annual maintenance at a realistic proportion of the build, plus the infrastructure. The comparison people usually run — licence against build — makes the package look free and is the reason so many implementations overrun.
What is a realistic maintenance figure for bespoke software?+
Plan for a meaningful annual proportion of the original build, and treat anything much lower in a proposal as optimism rather than efficiency. It covers dependency and security updates, platform changes forced from outside, and the changes the business asks for once it starts using the thing. A build quoted with no maintenance line is not cheaper; it has moved a cost to a year when nobody budgeted for it.
Can we start with a package and build later?+
Yes, and it is often the sensible sequence — the package teaches you what the process actually needs, which is the information the build required and did not have. Two conditions make it work: keep customisation minimal so the package stays upgradeable, and confirm at contract time that you can extract your data in a usable form. Without the second, "later" means a migration project nobody costed.
In this cluster
A short analysis that separates the processes worth keeping from the ones only kept out of habit, which is what actually decides this.
Find out which processes are load-bearing