Case study diagrams

The figures from both cases, at full size. Each one argues the turn its case rests on, so the short account here is the argument rather than a caption.

01 · Enrollment Framework

The convergence

Three requests reached the company through three different doors, each looking like a one-off build for the customer who asked. The figure holds them side by side to show what the intake routes hid: the same underlying capability, requested three ways. Below the convergence is what got built once instead of three times, and the program that came after it, set up rather than built.

Read the Enrollment Framework case →

The convergence. Three requests run across the top: request 1, contracted, arrived through an agreement; request 2, which looked like an enhancement, arrived through Support; request 3, framed as a compliance need, arrived via the CSM. Three lines drop from them and converge on a single point. Below, a panel titled The Program Enrollment Framework lists three steps: sign up, a person enrolls themselves or an object they own in a program; match, the system checks the enrollment against the official record to confirm it; handle differently, from then on defined events on that record get the treatment the program calls for. Beside the panel: we built it to be configured, so customer programs that come after go live faster.

02 · Accounting Escalations

The missing signal

Two tracks descend in parallel. The left one is what the ticket data could see, and it leads cleanly to a year of structural work. The right one never enters the dataset at all, because nobody opens a ticket to say they do not understand the screen in front of them. The figure exists to show that the analysis was not wrong about its evidence, it was wrong about what its evidence could contain.

Read the Accounting Escalations case →

The missing signal. Two columns descend in parallel. The left column, what the ticket data could see, runs from 170 escalations through AI pattern analysis to the conclusion that the reporting and transaction path is broken, ending in a large three-team refactor at 12 or more months, many dependencies and high cost. The right column, the unrecorded signal, runs from daily how does this work questions from support and frontline, through a superuser added to the support team, to ticket volume drops. Both columns curve into a single point, which leads to reframe the problem: not missing functionality but comprehension and fluency. The outcome below reads suspend refactor.