Build, Buy or Integrate?

The question behind the question

“Should we build or buy?” is really “is this process a commodity or a differentiator?” Accounting is a commodity: nobody wins on having unusual bookkeeping. The way a specialist insurer prices unusual risk might be the entire business.

Buy the commodities. Build the difference. The mistakes go both ways — businesses that build their own CRM, and businesses that force their genuinely unusual operation into a product designed for someone else's.

Buy when

• The process is standard across your industry
• A mature product covers 80%+ of the requirement
• Compliance is involved and someone else will maintain it — payroll, tax, accounting
• The total cost of ownership over three years is clearly lower
• You have no appetite for maintaining software long term

Accounting, payroll, email, HR records, helpdesk, document storage. We have never recommended building any of these, and we have talked several clients out of it.

Build when

• The process is genuinely how you compete
• No product fits without changing how you work in a way that costs you an advantage
• You need it to connect several systems in a way nobody else needs
• Licence costs at your scale exceed a build within about two years
• The data is sensitive enough that where it lives is a decision in itself

The test we use: if a competitor bought the same off-the-shelf product tomorrow, would you have lost anything? If not, buy it.

Integrate when both are true

The most common right answer, and the one most often missed. Keep the off-the-shelf product for the standard 80% and build the 20% that is yours around it.

In practice that means using the product's API rather than replacing it: a custom quoting engine that writes into a standard CRM, a bespoke scheduling layer over a standard calendar, a pricing service that feeds a standard e-commerce platform.

This is usually a quarter of the cost of building the whole thing and keeps you on the upgrade path of a product with a full-time team behind it.

Running the three-year numbers

The line people forget is licence cost growth. A per-seat product at £40/user/month is cheap at ten users and £48,000 a year at a hundred. If your headcount is growing fast, run that curve before deciding.

When we recommend buying, we say which product

By name, and without any reseller relationship or referral fee. We are asked often enough that we keep an informal list of what we have seen work in each category.

It costs us the project and it keeps the relationship. Several of the clients we have built substantial systems for first came to us with something we told them to buy instead.

Frequently asked questions