Architecture
Drift Index
Weighted average deviation severity across audits in the trailing 90 days.
deviationScore
across all audits in
trailing 90 days
What it signals
ADI is the depth signal that DCS cannot provide. Where DCS is binary — an audit either passed or it did not — ADI captures the degree of deviation. An audit that fails one minor Lens and an audit that fails across every Lens both count equally against DCS. ADI reflects severity and breadth of drift, not just its presence.
The clearest illustration: two orgs can have the same DCS of 75%. The one with ADI=0.15 has minor, contained violations — a few low-severity Lens failures concentrated in a corner of the codebase. The one with ADI=0.55 has deep structural divergence — major and blocker-severity failures spread across multiple systems. The headline score looks identical. The architectural reality is entirely different.
A DCS decline accompanied by rising ADI means violations are getting worse in severity, not just frequency. A DCS decline with flat or falling ADI means more audits are failing, but each failure is minor. ADI turns what would otherwise be a single-number conformance summary into a severity-aware architectural signal.
How rkito produces it
Each audit produces a deviationScore based on the severity and count of failing Lenses in that audit. ADI is the weighted average of that score across all audits in the trailing 90-day window. It surfaces on the org dashboard at /changecontrols/dashboard.
Who this metric is for
ADI is the depth signal DCS cannot provide. A DCS decline accompanied by rising ADI means violations are getting worse in severity, not just frequency. Without ADI, two architecturally different orgs look identical at 75% DCS. With it, one reads as minor contained drift and the other as structural emergency.
DCS tells you whether drift is happening. ADI tells you how bad it is. An ADI below 0.20 trending flat is a healthy architecture. An ADI above 0.40 trending upward is an escalation signal that warrants an architectural remediation project, not a routine sprint fix.
ADI surfaces which failing audits are trivially minor and which represent deep divergence. Use it to prioritise which FAIL-status audits to open CHIs for — sort by deviationScore descending at /changecontrols/designaudits to find the highest-severity drift first.
How this metric gets misread
ADI can be artificially low if most Lenses in the org are INFO or MINOR severity. An org with only low-severity Lenses will show low ADI regardless of architectural health — there is simply no high-severity signal to pull the score upward. A low ADI in this context is not evidence of good conformance; it is evidence that the Lens library has not been calibrated to catch serious violations.
Always confirm that BLOCKER and MAJOR Lenses exist alongside INFO and MINOR ones. An ADI of 0.05 with zero BLOCKER-severity Lenses configured is not a healthy architecture — it is an ungoverned one. Check Governance Depth (GD) and Lens severity distribution before reading ADI as a positive signal.
What to do at each threshold
Step-by-step response playbook for each signal state.
Drift is minor and contained. Confirm the governance foundation is real, not absent.
Confirm DCS is also high. Low ADI with low DCS indicates few audits are running, not good conformance. ADI cannot distinguish a well-governed org from one where almost no audits occur — DCS provides that context.
Verify BLOCKER and MAJOR Lenses are present in the active Lens library. A low ADI built entirely on INFO-severity Lenses is not a governance success — it is a calibration gap.
Material drift is present. Find the worst-severity violations before they compound.
Navigate to /changecontrols/designaudits, filter by FAIL, sort by deviationScore descending to identify the highest-drift entities.
Open CHIs for each material discrepancy. Distinguish real violations from Lens miscalibration — the former requires an awareness and remediation response, the latter a recalibration sprint via CHI.
Deep structural divergence. This is not a routine fix — treat it as an architectural remediation project.
Assign a lead architect, define scope, and track remediation via CHIs. An ADI above 0.40 indicates the architecture has diverged broadly enough that ad-hoc fixes will not converge it. A structured project is required.
This is not a routine fix. Escalate to architecture leadership. Define remediation milestones with measurable ADI targets — not a single sprint deliverable but a tracked programme with clear exit criteria.
The binary conformance rate ADI contextualises. DCS tells you how often drift occurs; ADI tells you how severe it is when it does.
The proportion of flagged violations that remain open. High ADI combined with high UDR signals drift that is both severe and unaddressed.
The rate at which architectural debt is accumulating versus being resolved. ABR tracks the remediation response to the drift ADI surfaces.
What is your org's ADI?
rkito calculates ADI automatically from your design audits. Connect your repo and see the true severity of your architectural drift within minutes.