Where self-build budgets actually go wrong: groundworks
Nobody overspends on kitchens by forty thousand pounds. They overspend below ground, before anything visible has been built.
Ask someone mid-build where the money went and they will usually say the same thing: it went into the ground. Substructure is the part of the job with the least visible output, the least design attention, and by some distance the widest range of possible outcomes.
That is not because groundworkers are unreasonable. It is because the foundation design depends on what the ground is actually made of, and on most self-build projects nobody has properly established that before the price was fixed.
The survey that pays for itself
A desk study and a modest ground investigation — a few trial pits or boreholes, some soil classification, a percolation test if you need drainage — is typically a low four-figure cost. It tells you the bearing capacity, the water table, whether there is made ground or contamination, and how deep you need to go.
Without it, your foundation is priced on an assumption. Assumptions are usually written into the contract as a provisional sum or an exclusion, which means the risk sits with you. The survey does not remove the risk from the ground — it moves it out of the unknown and into the budget, where you can do something about it.
Made ground and soft spots
Made ground is anything that was put there rather than formed there: old foundations, demolition rubble, backfilled ponds, farmyard hardstanding, general historic tipping. It is common on infill plots, which is exactly the kind of plot self-builders buy.
The problem is that it usually cannot carry a load, so you dig through it until you reach something that can. If that is 600mm down, you have a normal foundation. If it is 2.4m down, you are into deep trench fill, or piles and a ground beam, and the cost of the substructure has changed category rather than gone up a bit.
Trees and clay
On shrinkable clay, trees matter enormously. Clay shrinks as roots draw moisture out and swells when they stop, and both directions damage foundations. The required depth depends on the soil's plasticity, the species of tree, its mature height, and its distance from the building — and it is not intuitive. A single mature oak can dictate foundation depths across a surprising area.
The other half of the problem is heave: if a tree is removed, the clay reabsorbs moisture over years and swells upward. That is why a plot which has just been cleared of trees deserves more scrutiny, not less. Removing the tree did not remove the problem.
Sloping sites
Moving earth is expensive, and moving it off site is much more expensive than moving it around on site. A cut-and-fill exercise that balances — where the material you dig out is used to build up elsewhere — is dramatically cheaper than one that generates two hundred tonnes of surplus needing a waste transfer note and a lorry.
Slopes also bring retaining structures, which are engineered elements with real costs and real design fees, and stepped foundations, which take longer to build. If the site falls more than about a metre across the footprint, get the level survey done and the substructure properly designed before you fix a price with anyone.
Drainage
Foul drainage to a mains sewer is the simple case, and even then the depth and route matter — a connection that requires a deep dig in the highway is a Section 106 agreement and a road closure, not an afternoon.
Off mains, you need a treatment plant and somewhere for the treated effluent to go, usually a drainage field, which requires ground that percolates within a defined range. Ground that percolates too slowly fails; ground that percolates too fast also fails. Percolation testing before purchase is a modest cost against the possibility that the answer is no.
Surface water is the one people forget. Soakaways need the same percolation logic, and if the ground will not take it you are into attenuation and a discharge agreement, which is another cost and another consent.
How much contingency, and when to spend it
On a self-build with a proper ground investigation, ten per cent contingency is defensible. Without one, fifteen is closer to honest, and the majority of it should be regarded as belonging to the substructure until the foundations are complete and signed off.
The discipline that matters is not the percentage — it is refusing to spend the contingency on upgrades while the highest-risk part of the job is still in front of you. Contingency spent on better windows in month two is not available for two metres of unexpected made ground in month three.