A measured survey is a drawing decision you make standing in the building, not a pile of numbers you collect and decode back at your desk. That distinction is the entire difference between a survey that draws up in a day and one that sends you back across town with a laser measure and an apologetic email to the client.
The numbers are the easy part. A disto does those. The hard part is deciding on site what the drawing has to say, and then making sure every dimension you take can be checked against another dimension before you leave the building. Get that right and CAD becomes typing. Get it wrong and CAD becomes archaeology.
Here is the method, in the order you'd actually do it.
Draw the Building Before You Measure a Single Wall
Sketch first. Always. Freehand, roughly to proportion, big — one floor per A3 sheet, landscape, with at least a third of the page left empty around the plan for dimensions to live in.
You cannot decide what to measure until you've decided what you're drawing. Sketching forces you to look at the building as a plan rather than as a series of rooms, and it surfaces the awkward bits immediately: the chimney breast that isn't centred, the rear wall that steps out 400mm for no visible reason, the cupboard under the stairs that might actually be a bulkhead. Those are the things that bite you in CAD.
Use a fat pen. It stops you cramming forty figures into a corner you'll be unable to read on Thursday.
Outside-In, and Everything Off a Single Zero
Measure the envelope first, as a closed loop. Walk the outside of the building, take the overall length of each face corner to corner, and close the loop back where you started. That loop is your control. Every internal dimension you take afterwards gets checked against it.
Then inside, and this is the bit juniors get wrong: take running dimensions, not individual ones. Stand the disto in one corner of the room, treat that corner as zero, and record every feature along that wall as a cumulative figure from zero. Reveal at 1,240. Other reveal at 2,145. Chimney breast at 3,010. Far corner at 4,380. One string, one origin.
Chained dimensions — 1,240 then 905 then 865 — are a queue of small errors holding hands. Each reading carries the last one's mistake. Running dimensions can't accumulate because every figure is independently measured from the same point, and in CAD you type them as a single coordinate string rather than reconstructing a chain and discovering it lands 40mm past the wall.
Wall thicknesses come free at every opening, so take them there, and note what's included. A 350mm reveal might be a 215mm brick wall with 60mm of dry lining either side, and knowing which is which changes what you can do to it later.
Close Every Room or CAD Will Close It for You
Two diagonals in every single room, corner to corner, both ways. Not "a few diagonals where it looks dodgy". Every room.
Rooms are not rectangles. Victorian rooms particularly are not rectangles, and a room that's 20mm out of square at each corner will still look fine on your sketch and refuse to close in CAD. With four wall lengths and two diagonals you have a geometry that can be triangulated and checked. Without the diagonals you have four numbers and a hope.
Do the arithmetic on site if you can — or at least eyeball it. If the diagonals of a supposedly rectangular room differ by more than a few millimetres, the room is out of square and you now know that while you're still standing in it.
When furniture blocks the diagonal, take it in two parts to a marked intermediate point: corner to the far edge of a door reveal, then door reveal to the opposite corner, with the reveal noted on the sketch as the pivot. Two readings, still triangulated, still checkable.
One more at building scale: an overall external diagonal across the footprint. It catches the error that every individual room passed.
One Datum. Everything Else Is a Difference.
This is the single thing that separates a survey you can draw a section from and one you can't.
Pick one point in the building — usually the ground floor finished level in the hall or at the front threshold — and call it 0.000. Mark it with tape. Photograph it with something identifiable in shot so you can find it again. Then relate everything else in the building to that number as a plus or a minus.
If you have a laser level, set a line at 1,000mm above the datum and work off it, measuring down to floors and up to ceilings and heads. If you don't, run it manually with the disto and a straight edge, floor by floor, carrying the datum up the stairs by measuring each flight's total rise.
Every threshold. Every cill. Every step, half landing, hearth, ceiling, window head, door head, external ground level at each face of the building. In an old building the ground floor is rarely one level, and the rear addition is almost never at the same height as the main house.
A survey without a datum gives you plans. A survey with one gives you plans, sections and elevations that agree with each other. Skip it and the return visit is already booked, you just don't know it yet.
The Eight Things People Come Back Without
Straight from our own practice — these are the omissions that force a second visit, in rough order of how often they do it.
Structural span direction. Which way the joists run decides whether a wall can move. Read it off the floorboards, the ceiling line, a loft hatch, or the underside of the stair. Draw an arrow on your sketch in every room.
Reveal and cill depths. Reveal depth, cill projection, cill height above floor. You need all three to draw a window in plan and section, and you'll have taken none of them if you only measured openings as widths.
Ceiling voids. Ceiling height in every room, plus whether that ceiling is original or dropped. If the ceiling sits below the window head, something is hiding up there — a beam, a duct, a bad decision from 1987 — and it changes your section.
Soil and vent positions. The SVP, waste stub-outs, extract terminals, boiler flue. These dictate where a bathroom or kitchen can realistically go, and they're outside where you weren't looking.
Boundary and party wall lines. Which wall is the party wall, where the boundary sits relative to the building, where fences and gates land. Get this wrong and you've designed something you can't build.
Headroom under and over the stairs. Rise and going of each flight, number of risers, headroom at the landing and at the pinch point. Stairs are the most-redrawn element in any conversion.
Opening heights. Every door head and window head, not just widths. Without them your elevations are fiction.
Floor build-up at thresholds. The step between the hall and the kitchen isn't decoration, it's the difference between a suspended timber floor and a slab.
While you're at it: metre cupboards, consumer unit, stopcock, loft hatch size and position, radiators, manholes in the garden. None of them take thirty seconds and all of them get asked about later.
Photos Are Measurements, If You Take Them Properly
Number every photo on your sketch, at the position it was taken from, with an arrow showing which way you were pointing. A photo without that is a mood board.
Shoot every wall straight on, as square as the room allows, with a tape run flat along the wall and visible in shot. That way the photograph carries a scale and you can re-measure off it when you get back and find a figure you can't read. It takes an extra few seconds per room and it has saved me more return visits than any other habit on this list.
Then photograph your own sketch pages before you leave site. Notebooks get left on trains.
When the Wall Won't Let You Measure It
Sometimes the building says no. The party wall runs through next door's kitchen, the loft is boarded and full, the cupboard is padlocked. In practice nobody is granting you access rights over the fence on a Tuesday afternoon.
So derive what you can — party wall thickness inferred from the reveal at a window, an unmeasured return calculated from the external loop — and then flag it. We cloud it on the sketch and write ASSUMED next to it, and that note carries all the way through onto the issued drawing. A dimension you estimated and labelled is professional judgement. A dimension you estimated and didn't label is something a contractor will build.
Same principle for tolerance. If a wall is bowed, out of plumb, or you took the reading to a skirting rather than the plaster, write it down. If the room is occupied, work around it honestly: run your dimensions along the clearest wall, close with diagonals to reachable fixed points, and record the depth of whatever is in the way rather than guessing behind it.
Before You Pack Up: The Closing Loop
Don't leave until you've walked the building a second time with the sketch in your hand and done three things.
Add up the internal runs along each face and compare them to the external overall. If they disagree by more than the wall thicknesses account for, something is wrong and you are standing in the only place it can be fixed.
Check every room has two diagonals on the sketch.
Then walk through every doorway in the building and confirm you have both sides of it — the dimension in the room you're leaving and the dimension in the room you're entering, plus the reveal depth and the threshold level. Doorways are where surveys fail, because they belong to two rooms and end up measured in neither.
That loop takes about ten minutes and it is the actual difference between one visit and two.
Sometimes the Right Answer Is Not to Do It Yourself
Worth saying plainly: at Studio RBA we no longer send our own staff out to survey. On projects like the conversion we teach through, we insist on a measured building survey produced in CAD by an appropriate surveying company, and a topographical survey where there's site context. It's a decision about accuracy and about who carries the risk.
So if you're sent to a building that is genuinely beyond a disto and a notebook — multiple storeys, listed fabric, heavily obstructed, or nothing in it remotely orthogonal — say so before you go rather than after you've spent a day producing something nobody can rely on. A scan costs money once. A wrong survey costs a redesign, and then it costs the redesign again.
What the Survey Actually Becomes
Whatever lands on your desk — your own sketches or a surveyor's DWG — the first job is the same: clean it up. Join the line work, sort the layering, get the line weights consistent, trim the heavy external line at each opening and put a thin line in for the cills so the windows read properly. Surveyors draw literally; drawings need to be representative.
Then that cleaned file splits three ways. It stays as your existing drawings. A copy becomes the enablement set, with a thick red dashed line marking everything to be demolished and a note to the engineer where load-bearing status is unconfirmed. And the enablement drawing gets referenced under the proposals, so when the design moves, the demolition moves with it.
That chain — survey to existing to enablement to proposed — is exactly where The Project Guide Vol.1 picks up, starting with cleaning up a survey and running it all the way to a complete construction drawing package.
The survey is the only drawing on the job where you can't invent your way out of a gap. Everything downstream inherits it. Spend the extra forty minutes on site closing the loops, and you buy yourself a project that never has to argue with its own base drawing.
