Software gets bought like a subscription and lived with like a marriage. The evaluation is about features and monthly cost; the consequences are about what your business can and can't do for the next five years.
This isn't an argument against buying software. Most of what a business needs should be bought — accounting, payroll, email, anything regulated. It's an argument for evaluating it on the things that turn out to matter, which are rarely on the comparison page.
The four questions that matter more than price
1. Can I get my data out — all of it, on a schedule?
Not "is there an export button". Specifically: can another system read this data programmatically, without a person, including the custom fields we're going to add?
This determines whether the software becomes part of a joined-up business or an island. And islands are expensive in a way that never appears as a line item: every question spanning that system and another becomes a manual exercise, forever.
The answer also tells you something about the supplier's attitude. A system designed to be integrated is a system whose maker expects you to have other systems. One that's hard to read is often deliberately so.
2. What happens if we leave?
Every system will eventually be left. Ask now what that looks like: what format the data comes out in, whether historical records come with it, whether attachments and documents are included, and how long access lasts after cancellation.
The bad answers are worth knowing early. Data available only as PDFs. Attachments not included. Thirty days of access post-cancellation. None of those are dealbreakers on their own, but they're all a price you'll pay later, and they're negotiable at the start and never afterwards.
3. What does it cost at three times our size?
Per-user pricing that's comfortable at eight people can be uncomfortable at twenty-five, and per-transaction pricing that's negligible now scales exactly with your success.
The specific trap is software that becomes more expensive precisely as it becomes more embedded — by which point the cost of leaving exceeds the cost of paying. Model it at three times your current size before you sign, not after.
4. Who owns the workflow knowledge?
The subtlest one. As a business configures software, the software gradually becomes the documentation of how the business works. The rules, the stages, the categories, the automations — all of it accretes inside a system you don't control.
Leaving then means not just moving data but reconstructing years of accumulated decisions, most of which nobody has written down anywhere else. That's the real lock-in, and it's much stronger than any contract term.
Would we be comfortable if this supplier doubled its price, and could we act on it?
If the answer is no and no, you don't have a supplier — you have a dependency. That's sometimes an acceptable position, but it should be a conscious one.
When buying is obviously right
Regulated things: payroll, tax filing, accounts production, anything with a statutory format and a liability attached. The compliance burden is continuous, the consequences of getting it wrong are severe, and specialists carry that burden for a fraction of what it would cost you.
Commodity things: email, storage, calendars, video calls. There is no advantage available and no reason to think about it.
Things where the market is genuinely better: a mature CRM has had thousands of person-years poured into it. You will not beat that, and you shouldn't try.
When buying is the expensive answer
Three situations, and they're more common than the software industry suggests.
When the thing you need is specific to how you work. Off-the-shelf systems encode an average version of a process. If your process is genuinely a differentiator, adopting the average version to fit the software is a strange way to run a business — and the customisation required to avoid that usually costs more than building the narrow thing you actually needed.
When you only need 5% of it. Businesses routinely pay for large platforms to get one capability. The other 95% isn't free: it's configuration, training, and complexity everybody has to navigate forever.
When it needs to join things up. Almost no bought system is good at spanning your other systems, because it can't know what they are. The joining layer is the part that tends to be worth owning.
The hybrid that usually wins
The pattern that works for most small businesses isn't build-everything or buy-everything. It's: buy the regulated and commodity components, own the layer that joins them and the layer your customers touch.
Accounting stays bought. Payroll stays bought. Email stays bought. What you own is the data layer that reads across all of it, and the customer-facing pieces that are specific to you.
That combination is defensible for a long time. The bought components can be swapped without touching the layer above. The owned components carry your actual differentiation. And nothing critical depends on a supplier's roadmap.
The costs that don't appear in the price
Four of them, and together they frequently exceed the subscription.
Implementation. Configuration, data migration, and the internal hours spent on both. For anything operational this is routinely more than the first year's licence, and it's the number suppliers are most reluctant to quantify.
Training and turnover. Not just the initial rollout — every new starter, forever. A system that takes a day to learn costs a day per hire indefinitely.
The workaround tax. Where the software doesn't quite fit, people build compensating spreadsheets. Those become load-bearing, and they're a permanent cost caused by the software rather than solved by it.
Integration. If it can't be read programmatically, every question spanning it and another system becomes manual work forever. This is the cost that compounds hardest and gets discussed least.
Subscription × 60 months, at your projected headcount rather than today's, plus implementation, plus a realistic figure for the workaround spreadsheets. Then compare that with building the narrow thing you actually needed.
The comparison often surprises people in both directions — which is the point of doing it rather than assuming.
Questions that reveal how a supplier thinks
Beyond the four big ones, five that produce unusually informative answers.
"What are you bad at?" A supplier with a considered answer is being straight with you and knows their product. One who says "nothing" is either inexperienced or selling.
"Who shouldn't buy this?" Same principle, harder to deflect. Good software has a shape, and a supplier who knows it will describe the customer they'd turn away.
"What does the roadmap look like, and what happened to last year's?" The second half is the useful part. Roadmaps are aspirational; delivery against a previous one is evidence.
"Can I speak to a customer who left?" Nobody says yes. The reaction is still informative.
"What happens if we need something you don't do?" Establishes whether there's an extension route, a professional services conversation, or a flat no. All three are workable if known in advance.
The renewal is the real decision
Buying gets scrutiny. Renewal gets an invoice and a shrug, which is where most software waste lives.
Worth a calendar reminder sixty days before each material renewal, with three questions: is anyone still using it, has the price changed relative to what we get, and what would replacing it cost now versus a year ago.
The middle question catches the common pattern of per-seat pricing rising quietly with headcount while the value delivered stays flat. The third catches the case where a build has become cheap enough to be worth doing — which happens more often than it used to.
What to write into the decision
Before signing anything material, write down four lines: what data we can get out and how; what leaving looks like; what this costs at three times our size; and what knowledge will end up living inside it.
Half an hour, and it will change at least one decision a year. More usefully, it makes the decision reviewable — so when the renewal comes round in three years and the situation has changed, there's a record of what you thought you were buying.