The Coupling Matrix
Components on both axes, every dependency marked. The clusters are your real modules, whatever the org chart says — and where the two disagree is where your integration pain has been coming from all along.
A platform has seven components and four teams. Integration work is painful and everyone agrees the teams need to communicate better — there have been joint stand-ups for two years. What most reliably predicts which pairs of components are painful to change together?
Moving the boxes
Reorganising teams without changing the coupling moves the pain rather than removing it. The matrix is the thing to redraw first; the org chart is downstream of it, whatever the announcement says.
Stand-ups as a tax
Recurring cross-team ceremonies are usually the price of a boundary in the wrong place. They are worth paying only if you have decided not to move it — and mostly nobody has decided anything.
The component that touches everything
A row that is dense across the whole grid is not a component, it is a substrate. Owning it inside one consuming team means everybody else queues behind that team’s priorities.
It is not just systems
Draw the same grid with departments, or with the sign-offs a decision needs. The clusters are still real and still disagree with the reporting lines.
Draw this grid for anything you are about to reorganise, and draw it before the announcement. Two things that depend on each other both ways are one module, and splitting a module across two teams buys a permanent coordination cost that no amount of joint ceremony removes. Conway’s law is not a warning about culture — it is a statement that your architecture and your org chart will converge, and you get to choose which one leads.