Buying technology you will still want in five years

The standard platform selection produces a matrix. Requirements down one axis, vendors across the other, weighted scores in the cells. It is thorough, defensible and reassuring to a procurement committee, and it systematically fails to capture the two things that determine whether the decision was good.

The first is how much it will cost to change your mind about something in three years. The second is how much it will cost to leave.

Features are the easy part

Mature platforms in the same category converge on features. By the time you are comparing three credible options, the differences in the matrix are mostly differences in how each vendor describes the same capability. Scoring them to two decimal places produces precision, not accuracy.

What differs materially between them is the shape of the thing: what it assumes about your business, what it makes easy, what it makes possible-but-expensive, and what it quietly forbids. Those assumptions are where the friction will come from, and they do not appear in a feature list.

Questions that produce useful answers

• Show us the last three breaking changes you shipped and how customers found out.
• What is the most common thing customers ask for that you have decided not to build, and why?
• If we needed our data out in ninety days, what would that involve, and has any customer done it?
• Which of our requirements would you implement as a customisation, and what happens to it when you upgrade?
• Who at your company would be answering our support ticket at four in the afternoon in March?

These are uncomfortable, which is what makes them useful. A vendor who answers them straightforwardly is telling you something a reference call arranged by their own sales team cannot.

Cost of leaving

Exit cost is almost never modelled, and it is the single largest factor in whether a platform decision remains reversible. It has three components: getting the data out in a usable form, rebuilding the integrations, and retraining the people. The third is usually the largest and is almost always omitted.

A five-year total cost of ownership that does not include the cost of leaving in year five is not a total cost of ownership.

Reversibility as a criterion

We recommend scoring reversibility explicitly, alongside fit and cost. A slightly worse platform you can leave in six months is frequently a better decision than a slightly better one you cannot leave in three years — particularly in a category that is still moving, where the probability that your requirements change is close to one.

This is not an argument for indecision or for building everything yourself. It is an argument for pricing optionality honestly, and for noticing that the platform which scored highest on features is often the one that scored highest because it makes the most assumptions about how you work.