Levels are the only part of site analysis with a right answer. Sun paths can be drawn beautifully and still tell you nothing. Wind roses are decoration nine times out of ten. But the survey has numbers on it, those numbers are real, and they will decide your section whether you engage with them or not. Get them wrong and you find out at technical design, when the threshold is 400mm above the path and the only fix is a ramp nobody has room for.
So here is the sequence that actually works: read the drawing, walk the site with it in your hand, work out where the water goes, then pick a finished floor level. Everything else in the building follows from that one number.
A topographic survey gives you two different kinds of information and students routinely treat them as the same thing. They are not.
Spot levels are measured. Somebody stood there with an instrument and recorded that this exact point is at 42.37. That number is true. It is the closest thing to fact you will get from a piece of paper.
Contours are interpolated. The surveyor's software drew a smooth line between the measured points. Where the spot levels are dense — across a flat lawn, say — the contours are reliable. Where they are sparse, or where there's a hedge, or a steep bank the surveyor couldn't safely traverse, the contours are an educated guess dressed up as a line. A retaining wall a metre high can vanish entirely into a contour interval and reappear as a gentle slope.
So read the spot levels first. Find the highest, find the lowest, subtract. That is your total fall, and it is the single most useful number on the sheet. Then look at how the spot levels are distributed and ask where the surveyor was not. Those gaps are where your site visit earns its keep.
Check the contour interval in the key. At 1:500 you might be looking at a metre between contours; at 1:200, often 250mm. A site that looks dramatic can be nearly flat, and a site that looks gentle can be falling a metre across your proposed footprint.
Find the Datum or the Numbers Mean Nothing
Every level on that drawing is measured from something. Usually one of two things.
If the survey is tied to Ordnance Datum Newlyn, your levels are absolute — they relate to the national network and to the levels on the neighbour's survey, the drainage records and the flood mapping. Numbers in the 40s or 50s or 100s are a good clue.
If the surveyor established a temporary benchmark on site and called it 100.00, the levels are relative. Everything works internally but connects to nothing external. This is fine — most small projects run on a TBM — but you need to know which you've got, and you need to find the physical TBM. It'll be a nail in a tree, a mark on a manhole cover, a chiselled cross on a gatepost. The drawing tells you where it is. Photograph it. Every level you set out on site will be measured from that mark, and if it disappears under a skip the whole project loses its reference point.
Walk the Site With the Drawing, Not Afterwards
The site visit isn't a separate exercise from reading the survey. It's the verification pass. Print the survey at a sensible scale, take a pen, and go looking for the things the drawing can't tell you.
Lift the manhole covers. Or at least the ones you can. The cover level is probably on the survey; the invert level — the bottom of the pipe — often isn't, and it's the number that decides whether your building can drain by gravity. If your lowest floor level ends up below the invert of the only available connection, you're into pumped drainage, and that is a cost and a maintenance liability you want to discover now, not in a fortnight.
Look at what's already retaining. Old sleeper walls leaning out. A brick boundary wall with stepped cracking. A bank held up by nothing but ivy and optimism. These tell you the ground has been moved before, and they tell you the neighbour's levels are probably not what the contours suggest.
Find the low point and look for evidence of water. Standing water after rain, obviously. But also: silt lines on walls, rush and sedge growing where grass should be, a rotten fence base on one side only, a gravel drive with a scour channel through it. Water has been finding a route across this site for decades and it will keep using that route after you build.
Look for made ground. A flat terrace halfway up a slope didn't happen naturally. Bits of brick in the soil, a suspiciously level lawn, a change in vegetation — someone has tipped material here. Fill settles. Where you build on it matters enormously and it's a conversation for a geotechnical engineer, not for you and a pen.
Check your own levels. A laser level and a staff will let you verify half a dozen spot levels in twenty minutes. In practice we do this on every sloping site, because surveys get transposed, superseded and occasionally drawn from the wrong benchmark, and a 300mm error you find on site is a problem, while the same error found on site during groundworks is a claim.
The Arithmetic You Should Be Doing in Your Head
Gradient is fall divided by distance, expressed as 1 in something. A site falling 3.5m across 20m is 1 in 5.7 — steep, around 17%. That single ratio tells you more about what's buildable than a page of contours.
Cut and fill is just as blunt. Take your proposed footprint, work out roughly how deep you'd be cutting at the back and how much you'd be filling at the front, average the two, multiply by the area. Say a 10m by 8m footprint sitting on a slope where the back is 1.4m below existing ground and the front is 0.4m above: the cut averages out at something like 0.7m over the uphill half, so call it forty-odd cubic metres coming out. As a rule of thumb, excavated clay bulks up substantially once it's dug, so that's noticeably more lorry volume than the hole suggests, and every cubic metre leaving site costs money twice — once to dig, once to take away.
The design move this points at: balance your cut and fill on site if you can. Material that moves ten metres is nearly free. Material that goes to landfill is not. That one consideration, more than aesthetics, is why buildings on slopes so often sit half in and half out.
One more thing before you draw a cut. The Party Wall etc. Act 1996 brings in excavation close to a neighbour's structure, and on a sloping urban plot you will be cutting near a boundary almost by definition. Find out where their foundations are before you commit to a level, not after you've served notice.
Water Goes Downhill and Arrives at Your Threshold
On a flat site, drainage is a technical problem for later. On a slope, it's a design problem now.
The building you're about to insert will do two things to the existing water regime: it will block the path water currently takes, and it will add a large impermeable area at the top of that path. Whatever falls on your roof has to go somewhere, and on a sloping site "somewhere" is usually your own threshold, then your neighbour's garden.
So before the plan, sketch the water. Where does surface water enter the site, where does it currently leave, and what happens to that route when your building is in the way? If the answer is that you've created a dam across a natural valley, you need a land drain or a swale on the uphill side, and you need to find somewhere for it to discharge that isn't a boundary dispute.
The regulatory frame is Part H for drainage, and the planning-side expectation of a surface water hierarchy — infiltration first, watercourse next, sewer last — which is what makes soakaway testing worth doing early. On a clay slope, infiltration may simply not work, and finding that out at technical design rather than at pre-app changes your whole strategy.
Then the detail that catches people out: external paving must fall away from the building, and on a site that already falls towards it you have to build that fall in deliberately. We generally set paving to fall away at a shallow but unmistakable gradient, and where the slope makes that impossible, a channel drain across the threshold becomes non-negotiable rather than a nicety.
Three Ways to Sit a Building on a Slope
Step it. The building follows the ground in discrete level changes — split levels internally, stepped foundations below. Best on moderate slopes where the fall across the footprint is roughly the height of a flight of stairs or a half-flight. It gives you the most interesting sections and the most generous relationship to the ground on every side. Wrong when the gradient is gentle, because you'll create level changes that serve no purpose and wreck accessibility, and wrong when the client needs a single-level dwelling.
Plinth it. Build a level platform and sit the building on it, with fill or a podium making up the difference on the downhill side. Fast, simple, and the whole floor is at one level. Wrong when it turns the downhill elevation into a blank retaining wall two metres high — which is exactly what it does if you're not watching — and wrong on made ground, where the fill you're adding sits on ground that's already settling.
Cut it in. Dig into the slope and let the uphill side go partially below ground. Superb for thermal mass and for keeping the ridge height down, which matters more than anything else if the planning issue on your site is visual prominence. Wrong on a high water table, wrong when you can't get a gravity drainage connection from a floor that low, and wrong when the retained height starts demanding an engineered structure with tanking, drainage behind it and a maintenance regime. Retained height above roughly a metre is where the cost curve turns sharply upward, in our experience — below that you're building a garden wall, above it you're building a structure.
The Finished Floor Level Decides Everything Else
Pick the number last, and pick it knowing what it commits you to.
Raise the FFL and you gain gravity drainage, you get the traditional masonry between ground and damp-proof course, you push the downhill elevation further into the air and you make the entrance harder to reach level. Drop it and the approach gets easier, the building sits more quietly on the site, and you inherit a retaining problem, possibly a pumped drain, and a tanking detail.
That tension between keeping water out and getting a wheelchair in is the real design problem of a sloping site, and it resolves in the external works, not the building. The path has to rise to meet the door while the paving falls away from it. That's a contradiction you solve with a landing, a channel drain and a carefully graded approach — all of which need space, and all of which need to be in your plan from the start rather than added by someone else at Stage 4.
So: read the spot levels, not the contours. Find the datum. Lift the covers. Sketch the water before the plan. Then set the finished floor level and let the section fall out of it.
The slope isn't the constraint. Pretending it's a flat site with a bit of character is.
If you want the wider method around this — what else belongs in the analysis and how the pieces fit together — our ultimate guide to architectural site analysis covers the ground this article deliberately doesn't.
