The meeting audit
A quarter of engineering calendars, read like a ledger. The method, the categories, and the two findings that appear every single time.
Somewhere in most reviews we export a quarter of calendar data for an engineering org and read it the way an accountant reads a general ledger. It is the highest-yield afternoon in the diagnostic, and nothing about it requires a consultant, so here is the method in full.
Pull every recurring meeting touching engineers: name, cadence, duration, attendee count, and multiply into hours per quarter. Sort descending. The top of that table is always a surprise to its own participants, because recurrence launders cost: a weekly hour with fourteen people is 182 hours a quarter, a working month, spent without anyone ever approving a working month.
Then categorize, and three buckets cover nearly everything. Coordination (standups, planning, reviews of work in flight), decision (design reviews, incident reviews, anything that ends with a choice), and broadcast (all-hands, status readouts, demos). The ratios are the diagnosis. Healthy orgs we read run heavy on decision meetings and light on broadcast. Struggling ones invert it: hours of humans reciting status that a document could carry, while decisions happen in hallways afterward, unrecorded.
Two findings appear every single time. First, at least one recurring meeting survives its purpose: founded for a crisis or a launch, the occasion long past, attendance now a habit with a calendar hold. It is always the one nobody can name the owner of. Second, the same few senior engineers appear in a startling share of everything, because every meeting wants the person with context, and the org is converting its deepest context into its scarcest attention, one invite at a time.
The fixes are undramatic. Every recurring meeting gets an owner and an expiry date, renewal being a conscious act. Broadcast content moves to writing, with the meeting slot kept only for questions. And the overloaded seniors get a defended no, backed by whoever runs the org, because they cannot decline their way out alone.
The tradeoff: calendars are personal territory, and an audit reads as surveillance if it arrives as one. Do it in the open, publish the method with the results, and aim it only at recurring structures, never at individuals' days. The reversal condition is a small org that talks constantly and ships fine; some do, and their calendar ledger balances without intervention. Everyone else is spending working months by standing appointment. The ledger just has not been read.