Made-to-order configurator

The status, up front

The project never launched. As it was being handed to engineering, the company was moving to a new factory outside Moscow: a new building, new logistics, and part of the staff did not move. Covid on top of that. The front end stayed partial and the mechanics never reached production.

The reason has nothing to do with design, but it should be said first and by me.

The market problem

Glass is sold through sales managers. Competitor calculators do not actually calculate: you enter parameters, get an approximate price and still leave a phone number for a manual quote.

The task was a configurator that really calculates.

Getting to the knowledge

The domain knowledge lived in the client’s head. Two to four hours a day for several months he talked and I worked it out. Six months in, I knew glass at a technologist’s level.

There are literally thousands of constraints in this domain. Two examples make the point:

• a mirror cannot be tempered, the reflective coating fails in the furnace;
• a cutout too close to the bevelled edge will crack the panel.

Complexity measured by competence

An ordinary buyer gets a simple calculator: a rectangle, parameters, buy a full sheet or half, book a surveyor. Precision to the millimetre, and under the terms a wrong measurement is the buyer’s responsibility, which means the interface has to show that risk before the order is placed.

A professional gets a configurator built around flights and steps. A staircase is counted by flights, and inside a flight by steps: each one its own row with height, width and checkboxes for fixing and for a seam.

Seams are not decoration. A single sheet cannot give you a four-metre staircase without a joint, so the interface plans where the joint goes. Otherwise the factory decides, and not where you would want it.

Decisions worth showing

Glazing zones picked on an isometric drawing. Numbered areas on a drawing of the staircase instead of a form with fields. “Glaze from the second step to the fifth” cannot be described in words without error.

Two surveying conventions, by tread overhang and by riser. Surveyors were not retrained, the interface accepted both.

A survey sheet that can be drawn by hand and uploaded as a photo, signed by whoever is responsible on site.

Delegation by QR code with an electronic signature, so the buyer can pass rights to a contractor or an interior designer.

A designer’s commission inside the system. An interior designer sets their own percentage and produces the first estimate.

Multi-orders with different shipping dates: a glass panel on Thursday, a staircase in two months.

Hardware article numbers driving cutout selection. The cutout has to match the hardware exactly or the piece is scrap.

A post-order screen that states the timeline honestly: survey in a day, drawings in seven, production up to fourteen, installation up to ten. Plus the surveyor by name with a phone number and, on its own line, the authorised representative. That removes the argument about who was present at the survey.

Photography

Complex staircases are first built in cardboard and fitted on site, and only then made in glass. I shot the production myself, including from a drone.

Working with photographers was a separate job. Glass is hard to shoot: you have to show transparency, reflection, thickness and the edge at once. More time went into making glass look like glass than into the calculator.

What I would do differently

I would cut it into stages and insist on a narrow first release. Even a simple order dragged in the manager and surveyor back offices, and we went at every direction at once because the client wanted all of it at once. With staged delivery, part of the system would have shipped before the move.