Early access: the reports are free for now.Use code BRICKROADTEST at checkout.See prices

How it works

How this works

The whole method, including the parts we would like to be stronger. If you are going to trust a result, you should be able to see how it was made.

Points, not marks

Every topic is built on a set of points (we call them axes): how much structure you want, how you handle risk, how much consistency you can honestly keep up, how much you defer to experts. None of them has a right end. A tired parent on shifts is not wrong about screens. They have different limits, and the scoring measures fit, not merit. Nothing in the data can even express "better".

A sourced grid, with room in it

Each approach has a position on every point, a tolerated spread around it, and a source for both. Gentle parenting is wide on structure, because its own books disagree with each other, and narrow on punishment, where they do not. Without that spread, an approach with internal variety would be penalised for any answer away from its centre. That would be an artefact of the arithmetic, not a finding about you.

The questions adapt

The questions pick whatever resolves the most and stop when the ranking is stable, usually after nine to fourteen. Information alone would happily settle a result on four points, so coverage is a hard rule on top: a minimum number of points covered, and a few that must be asked outright. Every question describes a situation, never a doctrine, and no answer uses any camp's vocabulary.

A match, not a verdict

Each approach is scored on its own 0–100 scale, and the scores are deliberately not made to add up to anything. Two approaches can both fit you at 88, and that is true of them rather than a failure to decide. Forcing the scores to add up is what manufactures "87% gentle, 84% Montessori", and it would make your score depend on how many approaches happen to be in the grid.

"Not sure" is a real answer

It is not a skip. An unsure answer leaves the ranking alone and lowers the confidence of the points it touched. A point you never took a position on is left out of the scoring entirely. We never guess a position you declined to take, and the result names the points you left open rather than quietly filling them in.

The report is generated, not written by a model

Every sentence in the paid report comes from the grid and from text written in advance. No language model touches it. That makes it traceable (a disputed claim can be walked back to a cell and its source), repeatable, and free to produce, which is what lets the free result be complete rather than a teaser.

What this cannot do

It cannot tell you what to do with your particular child, and it does not try. It measures internal consistency, not real-world accuracy: the calibration runs simulated parents against the same grid that scores them, which catches broken arithmetic and proves nothing about whether the rows are right. Only a panel of real parents who already know what they do can measure that. Until one has, treat the result as words for something you already knew, not as a finding.

3 topics
40 approaches
98 questions
65 sourced references