How early can you tell a project will go over.

Agencies usually discover margin in arrears, when the invoice and the timesheet meet and someone explains the gap. Flow*form computes margin while the work happens: each project carries a fee, committed hours, and a blended rate, and the card recomputes from those three numbers. The fixture tenant keeps the numbers honest: Northwind's range launch at 33%, Verdant's radar at 44%, and exactly one project in the red, which is the point of the demo.

In the working month that fixture models, the Halcyon onboarding rebuild crossed to -3% in week 4. It had committed 320 hours against a 60,000 fee and burned 560 by then, so the card turned and Pulse carried it in the next morning digest. That is the instrument: not a report someone pulls, but a flag where you already look. The related playbook walks the same week-4 flag step by step with the live Flow and Pulse panels.

An early flag is a decision, not an alert to acknowledge. When the margin line bends, you have two honest moves while scope and price are still movable: cap the work or re-price it. The card gives the arithmetic; the digest ensures no one can say they did not see it. Teams that read margin once a week cut the surprise at month end, because the number stopped living in a spreadsheet someone forgot to open.