Preliminaries are a weekly cost, not a percentage
Price preliminaries as a percentage of the works and a programme that slips eight weeks is eight weeks of cost nobody recovers.
Preliminaries are the commonest place an estimate quietly loses money, and the reason is a method rather than a rate. Priced as a percentage of the measured works — eight per cent, twelve, whatever the last job came out at — they become a number that moves when the works move and stays still when the programme does. That is exactly backwards. The welfare unit, the site manager, the safety adviser, the accommodation, the utilities and the gate all cost the same per week whether the works are cheap or dear, and every one of them costs more when the job runs long.
A programme that slips eight weeks on a percentage-priced tender is eight weeks of welfare, supervision and hire that nobody has recovered and nobody can point at. The estimate was never wrong about the total; it was wrong about what the total depended on.
The right shape
A weekly rate, a duration, and a quantity. Two site managers for twenty weeks is a different number from one for forty, and both are computable from a programme rather than from last year’s percentage. When the duration changes the preliminaries change with it, which is the whole point — a re-priced programme should move the prelims automatically or the estimate is lying about its own sensitivity.
CONSTRUX prices every time-related head this way: preliminaries, site management, logistics, health and safety and quality each take a weekly rate and a number of weeks, and the percentage of works comes out as an output for benchmarking rather than going in as an input. That single reversal is what makes a tender answer the question "what happens if this takes six weeks longer" without anybody rebuilding it. The weekly rates sit in the same cost model as the measured heads, which is set out on how it works.
The failure it catches
There is a worse version of the same mistake, and it is commoner: excluding the time-related heads altogether. A blank box on a form is easy to leave blank, and an estimating tool that treats blank as excluded will happily produce a four-week job with nothing at all against welfare, supervision or site set-up. Somebody pays for those four weeks, and on a fixed price with them out of the offer that somebody is the contractor.
So a job of two weeks or more that prices not one time-related head is warned about by name. A genuine one-visit job exists and the warning is not a refusal — the estimator is the one who knows which this is. What the platform will not do is let a month of work go out priced as though nobody had to be there. A prolongation argument in adjudication turns on exactly this: what the time-related costs were per week, and whether anybody priced them in the first place.
See the record being made rather than described.
Launch the demo environment