- Work
- 06
- Northline
- 2024
The audit that arguedagainst hiring us
They came for a redesign. The interface felt dated. The audit found the design was largely sound and the library had been forked four times. We said so, in the room, and wrote a plan they could run without us.
- Audit
- Accessibility
- 2024
- $4,500
- No redesign
The brief
A redesign would have been a large invoice and would have decayed back to the same state within a year, because the mechanism that caused the drift was untouched.
Northline came in expecting to commission a redesign. The interface felt inconsistent and dated. The assumption, as it usually is, was that new design would fix it. We ran the audit first. That is the smallest engagement we sell, and it is how we keep ourselves from taking the wrong work.
Fig. 02
Impact against effort. Said in the room.
The redesign sits at the bottom of the list, where expensive work goes when the mechanism that caused the drift is still running. We put this on the table.
| Finding | Impact | Effort | Action |
|---|---|---|---|
| Name an owner | 9 | 2 | Do first |
| Contribution path | 8 | 3 | Do first |
| Deprecate the forks | 7 | 5 | This quarter |
| Token inventory | 5 | 6 | After ownership |
| Visual refresh | 3 | 7 | Not yet |
| Full redesign | 2 | 9 | Do not |
The approach
The review found the design work was largely sound. Contrast held. Type held. What was failing was governance: no contribution path, no owner of shared components, and four teams each maintaining a private fork of the same library.
We said so in the walkthrough, in the room, with the evidence on the table. The scatter in Fig. 02 is that conversation. Naming an owner and opening a contribution path sit high and cheap. A visual refresh sits in the middle. A full redesign sits bottom-right, where expensive work goes to die.
The recommendation was a governance model and a twelve-week internal remediation plan their own team could run. They ran it. We invoiced the audit. They referred three organisations in the following year.
Fig. 03
Four forks of one library.
Core last shipped fourteen months ago. Platform carries thirty-one local patches. Growth has a private button set. Data forked the date picker. Mobile is a fourth library, undocumented, in production. A redesign would have been eaten by this in a year.
Core
The published package
Last release 14 months ago.
Platform
31 local patches
Growth
Private button set
Data
Forked date picker
Mobile
Undocumented, in production
The outcome
$4,500
What the audit cost
Against a six-figure redesign that would have left the forks in place.
0
Redesign recommended
The design was not the failure. The mechanism around it was.
12 wk
Plan they ran themselves
Ownership, a contribution path, deprecation of the four forks. No studio retainers.
3
Referrals that followed
The honest no is how the next three yeses arrived.
Held to
Three laws from the canon, applied on this job rather than cited after it.
01
Cite the influence
The forks were named, with the teams who made them, in the room. Drift is not a mystery when you write down where each patch came from.
Read the law →02
Test with the assistive technology
The design held under a screen reader. The forks did not. Four date pickers, four keyboard maps. That is an access failure caused by governance, not by taste.
Read the law →03
A system, not a logo
They did not need a new look. They needed one library, one owner, and a path to contribute without forking. That is a system.
Read the law →
Next project
A telemetry product that stopped looking like a dashboard
Meridian · 2026
Something similar in front of you?
Describe the problem rather than the solution. We reply within two working days and will tell you honestly whether this is work we should take.