Website durability is not settled on the day a site goes live. It is settled several months earlier, when the tools are chosen, and nobody notices at the time.
A delivered site always works. The real question is whether it will still work, and above all still be modifiable, in five or ten years.
What makes a site unmodifiable
A site rarely stops working all at once. First it becomes hard to change, then expensive, then people give up. Failure is the last stage, not the first.
Three causes recur. A dependency on a third-party service that closes or changes its terms. A pile of plugins nobody can account for any more. And bespoke development whose author is unreachable, with no documentation.
In all three cases the site still works on the day the problem appears. That is what makes the diagnosis so late.
The replacement test
One simple test lets you assess a solution before committing to it. How many providers, within reasonable reach, could take this site over without rebuilding it?
If the answer is “only one”, the technology is not at fault: the dependency is. A common, slightly ordinary solution is maintained better than a brilliant one nobody else practises.
The test displeases providers, which is precisely the reason to apply it.
What ages well
Structured content ages well. Simple templates age well. Functions written legibly, with understandable names, age well.
What ages badly are visual effects tied to a fashion, integrations whose key is held by a third party, and anything built to impress on delivery day.
A sound rule is to ask, before each choice, whether it will need defending or explaining in three years. What is hard to defend gets replaced quickly.
Documentation, the only real guarantee
Documentation need not be long. It must say where the credentials are, what was installed and why, what was tried and abandoned, and what deadlines exist.
Written in an hour at handover, it saves whole days two years later. Written afterwards, it is never written at all.
It is the same reasoning applied to technical debt: what is deferred does not disappear, it accumulates in silence.
A choice that is paid for, or earned, at the end
A site built to last sometimes costs a little more to build. It costs considerably less to live with, and it is that second figure that weighs.
Website durability is therefore not a technical preoccupation. It is an economic decision, taken very early, by people who will only see its effects long afterwards.
The cost of taking over
One figure helps decide at the moment of choosing: taking over an existing site costs on average a third of what it cost to build, when takeover is possible at all. When it is not, everything must be rebuilt.
That gap between a third and the whole is decided at the moment of the technical choices, several years earlier. It is almost never put to the client then, because it does not appear on a quotation.
Asking about the cost of takeover before signing is therefore the only way of asking it at all.
It is also why some projects are designed in stages. WebHarmony works in this way in close collaboration with the Sageocracy association, whose platform is designed to accommodate, when the time comes, uses that do not yet exist. Laying an architecture today that can receive them costs little; rebuilding it later would cost everything.