Build or buy: a decision with downstream costs

The choice between in-house development and purchase is usually argued as a question of cost. It is above all a question of which dependency is bearable.

The decision to build a capability in-house or to buy it is usually justified with a cost comparison. That calculation is necessary, but it captures the smaller part of the decision.

What the cost calculation captures

On the purchase side stand licence or usage fees, integration effort and training. On the build side, staff, time to readiness and ongoing maintenance. These items can be estimated, and the estimate almost always favours purchase, because development effort is shared.

What it does not capture

What is not captured is how freedom of action changes over time. Buy, and you adopt the vendor's development path: their priorities, their timetable, their pricing. So long as those suit, that is an advantage. When they no longer suit, the effort required to change has already grown.

Build, and you keep the path but pay for it continuously — not only for the development, but for the knowledge that must be retained so the result stays maintainable. In-house work whose understanding rests with individual people is a dependency like any other, merely a less visible one.

The question of proximity to the core

A workable guideline: what distinguishes a company from others belongs in its own hands. What everyone needs in a similar form is a candidate for purchase. You buy accounting. The process that constitutes your own advantage, rather less so.

The difficulty lies in determining honestly what is genuinely distinguishing. The tendency to overestimate one's own share of it is widespread and expensive.

The middle way and its price

A middle path is often chosen: buy and adapt. That is sensible, but it has a cost centre of its own — every adaptation must be carried forward at each of the vendor's version changes. Anyone who does not budget for that effort has booked the middle way as the cheaper of the two models, when in fact it carries elements of both cost sides.