Measuring a practice that never measured itself.

One of the quieter findings of doing this work repeatedly is how little most legal organisations know about their own throughput. Not in a pejorative sense - there was never a reason to know. Firms recorded time for invoicing, which produces a number that answers a billing question and almost nothing else. Departments did not record it at all.

The consequence surfaces the moment anyone tries to demonstrate that a change worked. A practice that has automated its first-draft production will tell you, accurately, that things feel faster. Ask by how much and the honest answer is that nobody knows, because nobody measured the before. At that point the improvement exists in the experience of the people doing the work and nowhere else, which means it cannot be defended in a budget conversation, a partner meeting, or a pitch.

So the first fortnight of any engagement includes a baseline, and it is worth doing even if you never engage anyone. Four weeks of rough logging against named actions across every open matter of your highest-volume type. Not precision instrumentation - a low, a median, and a high per action is enough to price against and enough to prove movement against later.

The metrics that matter are usually not the ones that feel most natural. Hours saved is a weak claim, because it invites the immediate question of what happened to the hours, and in a firm that bills by the hour the answer is awkward. Cycle time from instruction to delivery is strong. Throughput at fixed headcount is strong. Revision rate on first drafts is strong, and it guards against the failure mode where speed improves because quality quietly dropped.

There is a second reason to build the loop beyond proving value to anyone else. A practice where every matter runs through the same named actions generates data about itself as a by-product - which actions overrun, where exceptions cluster, which matter types are mispriced. That data feeds back and the institution improves. Contrast the traditional model, where improvement accumulates inside individual lawyers' experience and walks out of the building when they leave. One compounds. The other resets, and it resets every few years without anyone recording the loss.