1. Billing and seats (do this first)
This section pays for the audit. Everything after it improves the portal; this part refunds it.
- Marketing contacts marked 'marketing' that haven't been emailed in 12 months — the single most common recoverable spend.
- Paid seats assigned to people who only need view-only access under the seat-based pricing model.
- Duplicate or overlapping add-ons carried over from an older contract.
- Hub editions you upgraded for one feature that a workflow could replace.
2. Data quality
- Duplicate contacts by email and companies by root domain.
- Records with no owner, and records owned by former employees.
- Free-text fields that should be dropdowns (country, industry, lead source).
- Contacts with no lifecycle stage, or stuck in a stage for over 180 days.
- Email validity — bounce rate and unengaged segments.
3. Data model and properties
- Count of custom properties, and how many have been written to in the last 90 days.
- Properties duplicating standard fields ('Company Name 2' is always there).
- Association labels — present, or everything associated flat?
- Whether the model supports the reports leadership asks for, or people export to spreadsheets.
4. Pipelines and lifecycle
- Stage definitions written down anywhere, and whether reps agree with them.
- Stages that represent internal admin rather than buyer progress.
- Deal rot: deals in stage longer than twice the median.
- Closed-lost reasons captured as a required dropdown.
- Whether MQL and SQL mean the same thing to marketing and sales.
5. Workflows and automation
If naming is the problem, our free workflow naming generator will produce a convention you can apply retroactively.
- Active workflows nobody can explain the purpose of.
- Workflows with no suppression list, no re-enrolment logic, or overlapping enrolment triggers.
- Automation that emails contacts based on stale property values.
- Naming conventions — a portal with 200 unnamed workflows is unmaintainable.
- Error notifications routed to a person who still works there.
6. Reporting
- Does one number (pipeline created, say) have two definitions in two dashboards?
- Reports built on properties that are only sometimes filled in.
- Attribution set up but ignored because nobody trusts the model.
- Dashboards with zero views in 90 days.
7. Integrations and AI readiness
- Sync direction and conflict rules for each connected system.
- Failed sync logs — check them, don't assume.
- Fields written by two systems, which is how data quietly corrupts.
- Whether AI features (Breeze summaries, prospecting and agent tooling) are drawing on properties that are actually populated. AI inherits your data quality; a portal with 40% blank job titles produces confidently wrong summaries.
8. Permissions and governance
- Super admins — how many, and should there be that many?
- Who can edit workflows and delete records in bulk.
- Sandbox usage before changes hit production.
- A named portal owner. If nobody owns it, the audit findings will still be true next year.
Common mistakes
- Auditing features instead of outcomes — nobody cares that you have 300 properties, they care that forecasting is wrong.
- Delivering findings without cost estimates or a sequencing plan.
- Fixing workflows before fixing the data they run on.
- Skipping the billing review because it feels unglamorous.
- Running the audit and then giving it to a team with no capacity to act on it.