The bathroom that ordered its own tiles

The project was a client apartment; the room was its bathroom. Nothing about either was unusual. What was unusual is what did not happen afterwards: no second tile order, no half-empty boxes stacked in the hallway waiting for someone to admit a mistake. This is a short account of why — written from our own studio, about our own tool, which you should weigh accordingly.

How tile quantities used to leave this office

The old method will be familiar. You open the model, turn on the tile texture, zoom the wall until it fills the screen, and start counting — full tiles by eye, cut tiles by squinting, the strip behind the door by hope. A calculator sits beside the keyboard; a waste allowance gets added on instinct. It is counting off a screen with a calculator and a prayer, and every studio we know has its own version of the ritual.

The order then arrives one of two ways. A box too few, which stops the tiler mid-wall and puts a batch-matching problem on your desk. Or a pallet too many, which the client pays for and stores in the cellar as a monument to caution. In our experience there is no third outcome — only the choice of which failure to buy.

The set-out

This time the set-out was run in DocSet, on the selected wall and floor faces, with the joint width the tiler actually intended to use — not a texture's decorative approximation of one. The distinction sounds pedantic until you remember that a joint repeated across a wall is a dimension, and dimensions accumulate.

The layout draws every cut piece, and that is the whole trick. It matters twice. Once at the drawing board: a sliver row under the ceiling or a mean cut on an exposed corner shows up while it can still be moved, so the argument about where the first tile sits happens in the model rather than on site. And once at ordering, because the count comes from the drawn layout — cut pieces included — rather than from an area divided by a tile.

A CSV the contractor could order from

The tile summary exports as a CSV, and we sent it as it came out. The contractor opened it in a spreadsheet, read it, and ordered from it — no transcription, no roughly, no phone call to establish what a hatch pattern had meant. When the delivery arrived, the boxes ran out at approximately the moment the walls did. We would like to report the exact margin, but we did not weigh the offcuts; the honest summary is that nobody drove to the builders' merchant twice.

What rode along for the rest of the flat

The bathroom order was the reason DocSet was open that week; it was not the reason it stayed open. The same model fed the rest of the flat's paperwork:

• The door schedule — dimensioned and exported as a CSV, no longer maintained by hand in a spreadsheet that was wrong the moment a door moved.
• Material areas by room and category — the flat's finishes summarised in one export, which is the table every quantity conversation with a contractor eventually arrives at.
• The product list — exported for pricing, edited outside the model, and re-imported without retyping a single row.

The sheet set itself started from DocSet's doc scenes, generated by level and sheet type. LayOut then began from a skeleton — named scenes in a predictable order — instead of the blank page and the scene housekeeping a blank page politely conceals.

What actually changed

We will resist the case-study convention of a percentage here, because we did not measure one and will not invent one. What changed is categorical rather than numerical: the tile quantity stopped being an estimate a person produced and became a property the model already had. The bathroom, in the only sense that matters, ordered its own tiles. We merely pressed export.