Nothing in a site analysis counts until it has changed a dimension. Not a diagram, not a page, not a lovingly rendered sun path with twelve arcs on it — a dimension. A building line pulled back. A cill height raised. A floor level stepped. If you can't point at a drawing and say this is here because of that, you did a survey, not an analysis.
So here is one plot, followed all the way through. A corner site in a 1930s outer suburb: two-storey end-of-terrace shop with a flat over, 9m to the high street, 17m down the side road, a fall of about 900mm from the back corner to the pavement. The client wanted a café at ground floor and two flats above. The numbers below are from that job. The sequence is the thing to copy.
The desk study is not a data-gathering exercise. It is a hunt for the things that will make the scheme illegal, impossible or expensive, before you've drawn anything you're attached to.
For this plot that meant: the planning history for the site and both neighbours (two refusals on the terrace, one for a dormer, one for a shopfront — both on design, both worth reading in full); the TPO register, because there is a mature lime on the verge outside the side elevation; the water company's public sewer map; the statutory undertakers' asset plans; the title plan, for easements; and the historic maps, mainly because a 1930s suburban corner shop usually has a tank or a forge somewhere in its past.
Two of those are worth dwelling on, because every competitor guide skips them. The public sewer map comes from the water company and takes a couple of weeks — in our experience the position of anything on it is approximate and the invert levels are optimistic. Private drainage inside your boundary is on nobody's map at all, which is why a CCTV survey is money well spent before you fix a ground floor level rather than after. And the title plan is where easements live. On this site it showed a right of drainage crossing the back third of the plot. That single line killed a basement plant room before the first sketch.
Budget-wise: for a student project, a day at the desk is enough and most of it is free. On a live job of this size we'd allow a day and a half of our own time plus fees for the arboriculturist, the CCTV survey and a measured survey. The analysis isn't the expensive part. Finding out in Stage 4 that you built over an easement is.
What Actually Got Recorded on the Visit
The visit was two hours on a Tuesday morning and forty minutes on a Saturday afternoon. Two visits, deliberately — a corner shop on a bus route is a different site at 8am on a weekday than it is at the weekend, and you only learn that by standing there twice.
What got written down, in a notebook, not a template: where people actually walked across the corner (not where the pavement suggests they should); the bus stop flag directly outside the high-street frontage and the queue that forms along it; the top deck of the double-decker, level with the first-floor windows and about four metres away; the wind coming up the side road hard enough to hold the shop door open; which gutters were stained and where the water went; the neighbour's blank gable; the only vehicle access, which is the side road, which is also the only elevation that gets decent afternoon sun.
Photographs: about ninety. Useful ones: maybe fifteen. The rule we teach is to photograph the problems, not the site — the stain, the crack, the queue, the puddle — because a photo of a nice brick wall tells you nothing in eight weeks' time.
Thirty-One Observations, Five Findings
The notebook came back with thirty-one things in it. Five made it into the design. The filter is one question asked of every observation: what would I draw differently if this were true? If the answer is "nothing", it's context, and context goes in the report as a sentence, not a diagram.
When two survivors disagree — and on this plot they did, badly — the weighting we use is blunt and has never let us down: legal beats expensive beats nice. A finding with a statutory or physical consequence outranks one with a cost consequence, which outranks one that would merely make the building better. Here, the best afternoon sun, the only vehicle access and the tree's root protection area all wanted the same corner. The tree won, because it was the only one of the three with a legal mechanism behind it.
The Five, and What Each One Moved
1. The queue. Roughly eleven hours a day, there are people standing still on the high-street frontage, facing the building. Move: the flat entrance came off that elevation entirely and went to the side road, because a door swinging into a bus queue is a door that gets propped. The café glazing stayed — a queue is an audience, and an audience is trade. Second move, in section: first-floor cills on the high street went up to 1.6m, so the bedroom floor sits above the sightline from the top deck. One dimension, one finding, directly traceable.
2. The 900mm fall.Move: the café floor was set flat and level with the pavement at the corner for a level threshold, and the back-of-house stepped up 450mm, landing on the natural ground at the rear. That step is now a 450mm plinth on the side elevation, which gives the brickwork a base to sit on and gives the back door a run of fall to the existing connection instead of a pumped station. One section move solved a threshold, a plinth and a drainage problem at once. If the arithmetic of this part is where you come unstuck, that's its own subject.
3. The lime and its root protection area.Move: the building line pulled back 2.8m on the south-west corner. That loss became the outdoor seating — which is exactly where the tree's canopy filters the afternoon sun that the elevation was going to be fighting anyway. The foundations on that side changed type to avoid excavation within the RPA. A constraint that cost floor area and bought the best room on the site, which is outside.
4. The drainage easement across the back third.Move: no basement, so the plant went to the roof, so the parapet rose 1.1m on the rear half, so the massing reads as two pieces rather than one. The most visible thing about the finished building was decided by a line on a title plan nobody had read.
5. The wind up the side road.Move: a lobby at the entrance, not a single door. And the cladding strategy inverted — the timber that was going to wrap the whole thing got confined to the sheltered north-east elevation, with brick on the south-west face taking the driving rain. Material specification, changed by forty seconds of standing in the wrong place with a notebook.
Five findings, nine drawn consequences. That is the ratio to aim for. If a finding produces no consequence it was never a finding.
The Two That Got Cut
This is the part the guides leave out, because it looks like failure. It isn't — it's the filter working.
The wind rose went in the bin. It came from a weather station some miles away and described a regional prevailing direction that had nothing to do with what the building would experience. The real wind behaviour was a street-level funnel between two terraces, and no rose at that scale could see it. The topic survived; the data source died. Worth noticing which one you're actually attached to.
The demographic study went too. Four handsome pages of census data on a catchment that the client's business plan had already defined, for a use that was fixed before we were appointed. It changed nothing. It would have padded the report nicely and we cut it, because every page that changes nothing trains the reader to skim the pages that do.
A views analysis nearly made it and shouldn't have. The only framed view on the plot was a church spire visible at an awkward angle from the back corner — which, given the plan, could only have been seen from a bathroom. Not every site has a view. Pretending otherwise is how you end up with a window justified by a diagram instead of a room.
The Audit Nobody Does
At the end of Stage 3, open the one-page findings sheet again and write a drawing reference next to each finding. Finding 1 — sheet 2.01, cill level. Finding 3 — sheet 1.02, building line. If a finding has no reference, one of two things is true: it was never a finding, or you've quietly lost it somewhere between the sketch and the coordination model. Both are worth knowing before the planning submission rather than after.
This is also the honest answer to the question of how you present analysis so it doesn't get shelved. It doesn't get shelved if it's one page and every line on it is answerable. Diagrams that survive the crit board work the same way — evidence or nothing.
Who to Ring, and When
Sequence matters more than the list. The arboriculturist comes before you fix the building line, not after you've drawn a corner you love. The CCTV and drainage survey come before you fix the ground floor level. If there's a shared wall, the party wall surveyor comes early enough that the notice periods under the Party Wall etc. Act 1996 — which run in months, not weeks — don't land on top of your start on site. The planning consultant comes last, once you have a massing worth testing. Ring them in the wrong order and you pay twice for the same advice.
And if you genuinely cannot get on site — restricted access, a competition deadline, a site on another continent — say so in the report, in writing, and list what you'd check first if you could. Street-level imagery from two different years will tell you a surprising amount about drainage, parking patterns and who maintains what. It will tell you nothing about the wind, the noise, or the queue. Which, on this plot, was four of the five findings.
