Gerald Ajam

Build as you go

Spend the first budget on finding out, not on finishing.

Status Position

Published September 2026

All notes

TL;DR

Most school spaces should be built in increments: a minimal first build, a short trial, and more spending only when real uses emerge. The exceptions are the things that must be done properly once.

The distinction it turns on

The usual reading

The capital project

Design the whole space, spend the whole budget, open it finished.

The argument here

The increment

Build what you are sure of, watch it used, then build the next part.

01The situation

The budget is spent before anyone teaches in the room

Most school space projects run the same way. A committee meets, a workshop produces a wish list, a vendor turns it into a design, and the whole budget goes on building it. The space opens finished.

Then Term 2 arrives. The features chosen in the workshop meet the lessons actually taught there. Some are used every day. Some are used once, for a visit. And there is no money left to change the ones that are not working.

02The tension

My interpretation

Nobody knows how a new space will be used until it is used

This is not a planning failure. A new kind of space is a guess about behaviour. Teachers do not yet know how they will teach in it, because they have not taught in anything like it. Students do not know either.

A capital project asks everyone to answer that question in advance and then pays for the answer in full.

A finished space is a bet placed before anyone has seen the game.

03The move

Find out first, then build

Building as you go replaces one big decision with several small ones, each made with more information than the last.

  1. Build first for the use you are sure of

    Start with the one or two uses you are confident of, and build only for those. For a large multi-purpose space, that might be discussion and collaboration for about fifty students in two zones, and nothing else yet.

  2. Borrow before you buy

    Trial furniture and portable technology for two weeks before writing a specification. Two weeks of real lessons tells you what a catalogue cannot: whether the tables move, whether the screen gets used, whether anyone sits in the soft corner.

  3. Keep what gets used, drop what does not

    Build, measure, learn, then scale. Extend what people use. Remove what they route around. Spend the next budget only when a clear use case has emerged, not when the next funding cycle opens.

  4. Hold a design long enough to judge it

    Keep a design for one to two years before redesigning it. Novelty inflates the first term, and a space judged on opening day is judged on its novelty. Iterating every term means never knowing what worked.

04Where it breaks

Some things must be done properly once

Increments work for furniture, zoning and portable technology. They fail for anything that is expensive to redo. Acoustics, light, ventilation, power and structural work cost far more to retrofit twice than to do well once, and they are also where the evidence on learning is strongest. Do these first, and do them properly.

There is a second risk. Increments can become an excuse for timidity: every step small and safe, nothing ambitious ever tried. Building as you go is a way of learning, not a way of avoiding decisions. If nothing in the first build could fail, it was not worth trialling.

05Where it lands

Budget for the trial, not just the build

The first budget buys information. The second buys what the information says to build. Most projects collapse the two and pay for certainty they do not have.

Separate what must be right first time from what can be learned into. Then spend on the first, and learn your way into the second.

The point

What transfers

The reason this page exists, rather than the story that produced it.

  1. Separate what must be right first time (air, light, sound, power) from what can be learned into (furniture, zoning, kit).
  2. Budget for the trial, not just the build. Two weeks of borrowed furniture is cheap insurance.
  3. Spend the next tranche only when a clear use case has appeared.
  4. Hold a design for a year before redesigning it. Judge it in Term 3, not on opening day.
Singapore