Lumin
Marketing Group

Custom applications / Lumin Marketing Group

Build or Buy: When a Custom Internal Tool Is Worth It

How to decide between another software subscription and a custom application, what each choice costs after the licence fee, and when to do neither yet.

The short answer: Buy when the process is standard and someone has already solved it properly: payroll, accounting, email, booking and most CRM work. Build when the process is the work your customers actually pay you for, when the subscription grows with every person you hire, or when the real system of record is a spreadsheet one person is allowed to touch. The licence fee rarely decides it. What decides it is who maintains the thing in two years and whether you can get your data out when you want to move.

Most of the time, buy

If a process runs the same way in your business as it does in ten thousand others, buying is the right answer and it is not a close call. Vendors have spent years on edge cases you have not met yet, and they carry the cost of security patches, tax table changes and browser updates. Buying properly also means finishing the setup, because plenty of tools get bought, half configured, then blamed for not fitting.

Building is worth considering when a standard tool models a process that is not the one you run, and a person fills the gap by hand every week.

Four questions that decide it

1. Is this the work you sell, or the work around it?

Invoicing is work around it. The way a restoration company sequences a job, or the way a clinic triages intake, is often the work itself. Standard software makes standard processes cheap, which also makes them identical to your competitors'. Fine for bookkeeping, expensive for the thing customers choose you for.

Take a property manager with a dozen buildings. Accounting belongs in accounting software and always will. But the way maintenance requests get triaged, assigned to a contractor and reported to owners may fit none of the products they have tried, and a coordinator spends part of every day copying between them. That gap is the candidate for a custom tool.

2. Does the price grow with your team or with your usage?

Per-seat and metered pricing behave differently as you grow. Microsoft licenses Power Apps per user, with the Premium plan listed at CAD $27.10 per user per month paid yearly and a free developer plan for building and testing. Zapier meters by task instead: its free plan includes 100 tasks a month and paid plans are sold in bands of tasks, so a workflow that fires on every enquiry costs more in a busy month.

Neither model is wrong. Twelve office staff on a per-seat tool is predictable. Forty field crew who each submit one form a day is where a custom build starts to pay for itself.

3. Can you get your data out on a bad day?

Read the export documentation before you commit, not after you want to leave. HubSpot, for example, documents its limits plainly: up to 300 exports in a rolling 24 hour window, up to three at a time, and download links that expire after 30 days. Those limits are workable, and the point is that you can find them before you sign.

4. Who is accountable for the personal information in it?

You are, whichever way you decide.

Alberta's Personal Information Protection Act sets the framework private sector organizations follow when they collect, use and disclose personal information, and gives people the right to see their own. Alberta, British Columbia and Quebec have private sector laws deemed substantially similar to PIPEDA, while the federal law still reaches parts of what you do, such as disclosing personal information across provincial borders. Where both apply, you comply with both, and PIPEDA's fair information principles begin with accountability.

Buying does not hand that accountability to the vendor, and building does not remove it. Only the questions change: with a vendor, where the data sits and who can see it; with a custom application, who set up the access controls, backups and audit record.

What building actually costs

A custom application is not the screen. It is the screen plus sign-in, permissions, backups, error handling, someone to call when it breaks, and a written description of how it works. A build that skips those is cheaper on day one and becomes the system nobody will touch by year two. So settle two things before anyone writes code: who owns this after launch, and what happens if that person leaves.

Internal assistants for research, drafting and operations add one more question, which is where the answers come from. One that drafts quotes from your approved pricing is useful. One that drafts them from whatever it can find is a problem you meet in front of a customer.

The middle path most businesses should take

The real choice is rarely build or buy in full. It is usually to keep the system of record you already pay for, finish configuring it, and build only the missing piece: the intake form your customers understand, the portal that shows a client their job status, the internal view that pulls four systems onto one screen.

That is a smaller project, and it keeps your data where your accountant and your CRM expect it. If the vendor ships that feature next year, you have not rebuilt the business on something you now have to maintain.

A checklist for this week

  1. Pick one process people complain about and write down every step by hand, including the copying between tools.
  2. Mark each step as standard or specific to you.
  3. Count how many people touch it and how often, then price the leading off-the-shelf option at that number and at your expected size in three years.
  4. Open the export documentation for every tool on the shortlist and confirm your own team can run it, with history, attachments and the links between records.
  5. List the personal information the process touches and where it would be stored.
  6. Name an owner for whatever you choose, bought or built.
  7. If you still want to build, scope the smallest version that removes the manual step, and agree how you will judge it six weeks on.

Where Lumin fits

  • We build custom applications for tasks existing tools do not handle well, and connected workflows so enquiries, email, calendars and internal tools need less manual copying.
  • We start from the business problem and use tools such as OpenAI, Claude, Gemini, local models and AWS where they fit.
  • If the answer is to buy something and configure it properly, that is a fine result of a scoping call.

Bring a repetitive task or a disconnected system to a discovery call and we will talk through scope, access and a practical first step.

Frequently asked questions

Is a custom application cheaper than a subscription?

Sometimes, and rarely in the first year. A subscription is a known monthly figure that includes maintenance. A custom build is a larger cost up front, a smaller running cost after, and an obligation to maintain it. It wins when per-seat or per-task pricing outgrows the value, or when the manual work filling the gap costs more than the build.

Can AI coding tools let my team build this themselves?

Drafting code is the part that got easier. Access control, data handling, backups, integrations and support did not. Small internal tools built this way can work when someone owns them. The common failure is a useful tool with no named owner and no record of how it works.

What should we never build ourselves?

Anything standard and regulated where mistakes are expensive and vendors compete hard: payroll, tax, payment processing, accounting and sign-in systems. Buy those, and spend the build budget on the work only your business does.