A plot eleven metres wide and twenty-eight deep, running back off a service lane behind a Victorian terrace. Former car repair workshop. Asbestos cement roof, oil-stained slab, brick boundary walls on three sides, one mature sycamore in next door's garden leaning over the back third. The client wants two bedrooms and a workspace. You have a week before the first design review.
That's the plot this article carries end to end. By Friday the analysis has to produce three sentences that start with the word because — because the sewer invert is where it is, the floor sits here; because of that window, the two-storey element goes there. If your analysis can't produce those sentences, it isn't analysis. It's a mood board with contours on it.
Everyone tells you to combine desktop and on-site research. Nobody tells you the order matters enormously. Desk first, always — because the site visit is the only part you can't repeat cheaply, and you want to arrive with questions rather than a camera.
What the desk actually gives you on a UK plot: Ordnance Survey mapping for the base, the local authority's planning portal for every application on and around the plot (including the refusals — read the officer's report, not just the decision notice), the Environment Agency's flood maps for surface water and fluvial risk, the British Geological Survey's online geology viewer for what you're building on, and historic OS editions for what was there before. For our plot the historic maps earn their keep immediately: a smithy in 1894, a garage from the 1930s. That's a contamination question before you've left the office.
Tree Preservation Orders: ask the council. The sycamore may be protected, may not be, and "mature trees may be protected" is the kind of sentence that gets repeated on every site analysis article without anyone checking. TPO status is a specific legal fact about a specific tree, held on a register you can ask for. Conservation area status brings its own notification requirement. Find out which applies before you draw anything within the canopy.
Print the OS extract at 1:200 on A3 and take it with you. The visit is for marking up that sheet, not for generating fresh atmosphere.
In our experience a plot this size takes about half a day at the desk, half a day on site and half a day drawing it up. That's a live small-project budget, not a student exercise — if you're spending a fortnight on a plot with three boundary walls and one tree, you're decorating.
Five Findings Move a Building. The Rest Is Decoration.
Here's what the ten-step listicles won't tell you: on a constrained urban plot, most of the categories they list don't change the drawing. Prevailing wind doesn't reorganise a two-storey house wedged between brick walls. "Cultural context" rarely relocates a staircase. Record them, by all means, but know they're context and not constraint.
Five things change the building. On this plot they came out like this.
Access. The lane is 3.2 metres between walls with no turning head. A rigid delivery lorry isn't coming down it and neither is a fire appliance, which pushes you into Approved Document B's hose-laying distances measured from wherever the appliance can actually stop. Bins have to be dragged roughly twenty-five metres to the street on collection day, and that run can't go through the living space. Construction access is the quiet one: a plot you can only reach through a 3.2m lane is a plot built in small components.
Levels. The slab falls about 1.4 metres from the rear boundary to the lane threshold. Not a slope you'd call dramatic, but enough that "ground floor" is now a question rather than a given. Every subsequent decision — threshold, drainage, retaining, where the steps go — hangs off one number you have to fix. (We've written separately on sitting a building on a slope; don't fix a finished floor level without reading that first.)
Daylight obstruction. Three-storey flank wall of the corner shop to the south-west. On site at 1pm in March, the front third of the plot was in full shade and the back two-thirds weren't. That's a free, honest measurement you cannot get from a sun path diagram alone, because the diagram doesn't know how tall next door is unless you tell it. Running the other way: the neighbour's rear ground-floor window faces your plot across the north boundary, and BRE's Site Layout Planning for Daylight and Sunlight sets out the 25-degree vertical test from the centre of that existing window. Draw that line on your section on day one.
Boundary and overlooking. Two first-floor rear windows looking straight down the length of the plot from about nine metres. Their garden runs along your east wall. Anything you glaze above standing height on that flank is a planning objection with a name and an address attached.
Drainage and ground. The public sewer runs under the lane. The invert — the bottom of the pipe, the level your drain has to reach — sits roughly 1.1 metres below the lane surface according to the sewer record. BGS says London Clay with made ground over it. Clay plus a sycamore plus a former garage slab is three separate foundation conversations.
That's the list. Five findings, each with a number or a direction attached. The sensory stuff — the lane smells of the chip shop, the rear is surprisingly quiet — goes in the report because it affects where you put a window seat. It does not outrank an invert level.
Test Findings Against Something You've Actually Drawn
This is the step that goes missing. Students produce beautiful analysis and then start designing from a blank sheet, as though the two activities happen in different buildings. The findings only become useful when there's a massing option in front of them to break.
Draw three, roughly, in half an hour. Cardboard or SketchUp, doesn't matter.
Option A — single storey across the full plot, courtyard in the middle. Daylight: fine, nothing obstructs anyone. Overlooking: solved by default. Area: short of the brief by about a third.
Option B — two storeys at the rear, single storey to the front. Area: works. But the 25-degree line from the neighbour's window, drawn properly in section, cuts straight through the rear first floor. Fails.
Option C — two storeys at the front where the corner shop already shades everything, stepping to single storey at the rear. Area: works. Daylight to neighbours: clears the 25-degree line. Daylight to us: the front rooms are in the shop's shadow, so the two-storey block gets bedrooms and the single-storey rear gets the living space and the workshop.
Notice what happened. The shadow from the shop stopped being a problem and became the reason bedrooms are at the front. That's the whole move. A constraint you can't remove is a programme instruction in disguise.
The Invert Level That Killed the Courtyard
Option C had a better-looking sibling. Keep the single-storey rear wing but drop its floor 700mm, giving a three-metre internal height inside a low external volume, with a sunken courtyard against the back wall. It drew beautifully. The section looked like something.
Then you put the drainage on it.
Floor at minus 700 means the rear gully sits somewhere around minus 850. The sewer invert in the lane is at roughly minus 1.1 metres, twenty-six metres away. As a rule of thumb we run 100mm foul drainage at 1:40, which eats 650mm over that distance before you've added a single bend. The arithmetic doesn't close. You'd be pumping, which means a chamber, a power supply, a maintenance contract and a client who rings you when it fails in year four.
And the excavation sits inside the sycamore's root protection area, which BS 5837 calculates from the trunk diameter — on a tree that size it reaches well past the boundary. Digging a courtyard in shrinkable clay next to a mature tree is a heave problem and an arboricultural report and possibly a refusal.
So the sunken courtyard died on Wednesday afternoon, which is exactly when you want it to die. Not at planning. Not at tender. The whole point of doing the invert level and the geology before you fall in love with a section is that the kill is cheap.
Write the kill down. "Sunken rear wing — ruled out, sewer invert at approx. -1.1m in lane, gravity drainage won't reach; also within RPA of protected sycamore." That sentence is worth more at your review than six pages of sun path.
Three Decisions, Stated Out Loud
End the analysis with a short list in plain language. No diagrams, no colour. Just this:
Finished floor level is set at the lane threshold plus 150mm, which fixes the rear as a stepped slab and gives gravity drainage to the sewer with fall to spare. Because the invert is at -1.1m and we're not pumping.
The two-storey element goes to the front of the plot, in the corner shop's shadow; the rear stays single storey.Because the 25-degree line from the neighbour's ground-floor rear window makes a rear first floor undefendable, and because shaded space is better spent on bedrooms than on a living room.
The east flank gets rooflights, not windows.Because their garden runs along that wall and anything glazed above 1.7 metres is an objection. The rear wing therefore gets a north-facing monitor roof, which is a section decision made entirely by a boundary.
Three sentences. Each one traceable to a finding, each one a thing you'd have to defend at a planning meeting. That's the deliverable. Everything else — the layered base map, the diagrams, the Design and Access Statement narrative — is the presentation of these three sentences to different audiences.
When the Desk and the Site Disagree, the Site Wins
On this plot, three things: the OS base shows a boundary wall that's been demolished and replaced with a fence 400mm inside the line. The planning portal shows the neighbour with consent for a single-storey rear extension; what's built is clearly two storeys with a flat roof, which means either an amendment nobody scanned or an enforcement issue you'd want to know about. And the sewer record shows a manhole where there is now tarmac.
None of those are unusual. Records describe intent; the site describes what happened. Photograph every discrepancy, mark it on the printed plan, and flag it in the report as unverified. A topographical survey and a CCTV drainage survey resolve the ones that matter — and on a live job you commission both before you spend fees on design, not after.
If you genuinely can't get on site — no access, hazardous, locked — say so explicitly and list what that means you don't know. An analysis that admits three unknowns is more useful than one that quietly guesses at them.
What a Bad One Looks Like
You can spot a superficial site analysis in about four seconds. Every diagram covers the whole site at the same level of attention. There's a wind rose on an enclosed urban plot. The sun path is geometrically perfect and takes no account of the three-storey wall next door. Nothing has a number on it. And critically — nothing was ruled out. A site analysis in which no idea died is a site analysis that wasn't tested against anything.
The fix isn't more research. It's drawing a massing option early enough that your findings have something to break.
Go back to the plot once more before the review, after you've made the three decisions, and photograph the specific things your decisions depend on: the neighbour's window from your boundary, the lane width with a tape in shot, the manhole cover. Those photographs are your argument. The rest were reconnaissance.
