What custom software actually costs

The number, first

Most projects here land around $10,000. Larger platform work — the kind that runs for a year and replaces an organisation’s core operations — is a different conversation and is scoped accordingly. If your budget is well under that, say so in the first call rather than the third; sometimes the honest answer is that an off-the-shelf product fits.

That figure is worth almost nothing on its own, which is why most sites decline to print one. The same sentence — ‘we need a member portal’ — describes a fortnight of work and a six-month build, and the difference is not visible until someone asks what a member is, who else can see their record, and what happens when one leaves.

So the useful question is not ‘what does it cost’ but ‘what makes it cost that’. Three things do most of the work, and you have more control over all three than a quote makes it look.

One: how many kinds of user there are

Cost tracks roles far more than it tracks pages. A site where everyone sees the same thing is a content problem. A system where a volunteer sees one subset, a coordinator sees another, an administrator sees everything and a client sees only their own record is an access-control problem, and access control is where the hours go.

Every additional role multiplies the states that must be designed, built and — the part people forget — tested. Two roles is not twice one role; it is one role plus every interaction between them.

If you want to reduce cost before you have spoken to anyone, this is the lever. Ask whether a role is genuinely distinct or whether it is one role with a different report at the end.

Two: whether data has to move

A system starting from nothing is cheap. A system that must absorb eleven years of spreadsheets, a legacy database and an inbox is not, and the expense is not the import script. It is the discovery that two departments recorded dates differently, that a third of the records have no unique identifier, and that some rows are duplicates that must be merged rather than deleted because both halves contain something real.

Migration is where honest projects overrun and dishonest ones quietly drop records. Ask any prospective developer when they will first load your real data. If the answer is ‘at the end’, the number they have quoted is a guess.

Three: what happens if it is wrong

This is the multiplier nobody mentions. Software where a mistake means a confusing afternoon costs one thing. Software where a mistake means an incorrect continuing-education certificate, a safeguarding record visible to the wrong person, or a donation receipted twice costs considerably more, because the work is no longer just building the feature. It is proving the feature cannot fail quietly.

For MP Education that meant credit tracking that stands up to an audit rather than a video player with a quiz on the end. For a case management platform holding sensitive family records it meant encryption and access rules designed before the screens were. Neither is a feature you can add afterwards at a discount.

The costs that arrive after launch

A quote covers building the thing. Running it is separate and permanent: hosting, a domain, email delivery, backups, dependency updates, and the security patches that arrive whether or not you have budget that month.

None of these are large. All of them are ongoing, and a proposal that does not mention them is either inexperienced or hoping you will not ask. Ask what the annual figure is and what happens to it if your usage doubles.

Where the money actually gets wasted

Not on the expensive-sounding parts. It gets wasted on features specified in a meeting, built faithfully, and never used — because nobody asked the person who would have used them.

The cheapest projects are not the ones with the smallest scope. They are the ones where the scope was decided by watching how the work is done today, rather than by listing what a system ought to have.

What the bands actually mean

Under $5,000 buys a site, not a system — and that is frequently the right purchase. Between $5,000 and $15,000 is where most of the work on this site sits: a real application, one or two user roles, data that matters, built once and maintained. Above that you are usually buying either a genuine platform or a migration with a decade of history in it.

If a quote sits far below the band its scope implies, the gap is being paid for somewhere — usually in testing, migration, or the assumption that you will not ask for changes. That is worth naming out loud before you sign, not after.

How to brief so the quotes mean something

Write down the three tasks that consume the most staff time today, who does each one, and what breaks when it goes wrong. That is a brief. It is more useful than a feature list, and it lets two developers quote the same thing — which is the only way their numbers can be compared.

Then ask each of them what they would leave out. The answer tells you whether they are selling you software or selling you a solution to the problem you described.