Step 1 of 12

Score the practice

You’ll go through 11 areas and pick a level from 1 (Partial) to 4 (Adaptive) for each. Be honest about what you see. This is your view only. Before we call it a formal practice score, someone still needs to sample real work.

How to pick a level

These rules keep us from talking ourselves up. A tidy average with one serious hole still fails people in production.

Score from the biggest real weakness. Don’t average up and hide a serious hole.
Don’t claim Level 3 unless you’ve seen real work items with the checkpoints and proof in place.
If you’re unsure, or people disagree, go with the lower score.
Learning across clients only counts for Level 4 if the fix is written back into the playbook.

What the levels mean

Level 1 · PartialIt works if the right person is around. Proof is thin, and quality depends on which team or client you’re on.

Level 2 · Risk-informedWe’ve got shared templates and risk calls, but people don’t always follow them.

Level 3 · RepeatableThere are required checkpoints, people know their roles, we keep proper proof, and we check the work regularly.

Level 4 · AdaptiveUseful checks run automatically, we spot trouble early, and what we learn on one client gets written back into the playbook for everyone.

Glossary (acronyms and terms)

RACI: Responsible, Accountable, Consulted, Informed. A simple chart of who does the work, who owns the call, who gets asked, and who just needs to know.
DoR (Definition of Ready): The checklist a request must meet before we start building. If it’s not ready, it doesn’t come into the sprint or build queue.
Risk tier: How risky the work is, from Tier 1 (lowest) to Tier 4 (highest). Higher tiers need stronger checks. Anything going to customers in production is at least Tier 2.
Stage gate: A required checkpoint before work moves on, e.g. design review or go/no-go.
QA (Quality Assurance): Someone other than the builder checks the work before it goes live. Self-checking alone doesn’t count as QA here.
Maker-checker / four-eyes: The person who built it can’t be the only final checker or the only person who signs off the release.
RCA (Root Cause Analysis): Working out why something broke so we can stop it happening again, not just patching the symptom.
DPIA (Data Protection Impact Assessment): A privacy risk review when handling personal data could hit people hard if we get it wrong.
Contact Key / Subscriber Key: The stable ID for a person in Marketing Cloud Engagement, so we don’t mix people up across channels.
Suppression: Rules that stop us messaging people who shouldn’t get the send (unsubscribed, opted out, and so on).
Evidence pack: The saved proof for a change: approvals, tests, checklists, and related records kept with the work item.
KPI (Key Performance Indicator): A number we track to see if quality or process health is actually improving.
SLA (Service Level Agreement): An agreed target for service quality, e.g. how accurate data should stay or how quickly we handle issues.
Assessment cycle: The label for this scoring round. Usually half a year: 2026-H1 is Jan to Jun, 2026-H2 is Jul to Dec.

About you

Your nameSo we know who responded when we look at the results.Role (optional)e.g. QA Lead, Solution Architect, Campaign Manager.Assessment cycleUsually the half-year label (H1 = Jan to Jun, H2 = Jul to Dec), unless you’ve been asked to use something else.

Loading assessment…

How to pick a level

These rules keep us from talking ourselves up. A tidy average with one serious hole still fails people in production.

Score from the biggest real weakness. Don’t average up and hide a serious hole.

Don’t claim Level 3 unless you’ve seen real work items with the checkpoints and proof in place.

If you’re unsure, or people disagree, go with the lower score.