On this page
A portal audit should leave the next person able to explain what needs attention and why. A score can help organise that conversation, but it is not a substitute for the evidence behind it. Start with a bounded question: is this portal ready for a new implementation, is an inherited automation still appropriate, or can a reporting definition be trusted?
This guide describes a review process, not a promise that every issue can be found automatically. Jetstack AI’s Portal Audit is read-only: it gathers evidence, scores supported checks and produces a report, including a branded PDF export. It does not repair the portal. Moving a finding into a Project is a separate scoping and approval decision.
1. Establish the boundary
Record the portal identity, the business process under review, the reviewer and the date of collection. Confirm that the connected account can read the relevant areas. HubSpot subscriptions, permissions, connected scopes and the available collection path can limit coverage. A missing value might mean unavailable access, not an empty portal.
Ask the operating team which processes must not be interrupted. Identify external systems that set properties, control ownership or send messages. A workflow that looks redundant in isolation may be a deliberate integration boundary. Keep an explicit list of exclusions so an audit of marketing automation is not presented as an assessment of the whole business.
2. Inspect the schema before the automation
Start with objects, property definitions and pipeline stages. Compare internal names, types, allowed values and business meaning. Two properties labelled “Region” might belong to different objects or encode different routing rules. A rarely populated property may still be essential to a seasonal process.
Then inspect the consumers: lists, forms, workflows and reports. Follow a field from its source to its final use. Look for a changed option value, an obsolete pipeline stage or a list rule whose meaning no longer matches the label. Do not delete or merge anything during this review.
For a concrete example, trace a contact’s service_region from a form through a list into an assignment workflow. Inspect whether the workflow’s enrollment, re-enrollment and suppression rules agree with the intended routing policy. Check where the value is populated by an integration as well as where people edit it manually.
3. Write a finding someone can act on
Use a stable record rather than an isolated screenshot. The following is an illustrative register, not a customer result or a Jetstack API payload:
finding: ROUTING-REGIONportal: Example Service Portalasset: Contact routing workflowobserved: Branch expects an option absent from service_regionevidence: Property definition + workflow branch inspectioncoverage: Definitions reviewed; enrollment behaviour not testedimpact_to_confirm: Contacts may miss the intended assignment branchowner: CRM administratordecision: Review intended option values before proposing a changeSeparate the observed condition from its possible impact. “The branch references a missing option” is an observation. “Leads are being lost” requires separate evidence. Attach the asset reference and collection context so a colleague can reproduce the review without guessing which similarly named workflow you meant.
4. Prioritise by exposure and dependencies
Review issues that affect active processes before unused experiments, but confirm that status with the team. Consider who is affected, whether the issue can propagate through dependent assets, and how reversible a proposed change would be. A formatting inconsistency and an unresolved ownership rule should not carry the same treatment simply because both are easy to count.
Group related findings under a decision. If a property option is inconsistent across a form, list and workflow, do not create three unrelated tasks that can be approved independently. Describe the shared dependency and the coordinated review needed. Where evidence is incomplete, the next task may be investigation rather than implementation.
5. Review the report, not just the score
Read the section evidence and coverage notes alongside the overall result. Confirm which checks were scored, which were not measured and which depend on context the portal cannot reveal. The scoring documentation explains the scoring model; it does not turn a score into a guarantee about business outcomes.
Before sharing a branded report, check the portal identity, selected content and any sensitive details appropriate to the intended audience. Follow the export documentation for the actual PDF workflow. A branded report is a presentation of the review, not independent certification of the portal.
6. Turn a finding into a separate work order
For each accepted finding, write the intended behaviour, the assets in scope, the proposed owner and the acceptance test. Get agreement on the change before authorising generation or deployment. In Projects, a reviewed plan precedes generation; deployment has its own approval. An audit finding is not permission to change HubSpot.
After implementation, inspect the relevant asset and run an appropriately controlled behavioural check. Record what was verified and what remains unknown. A successful API response or an improved score alone does not establish that the operating process now works as intended.
Further reading
Use the workflow audit article for a narrower automation review. Continue with the implementation guide to turn agreed findings into a bounded brief, or explore Portal Audit for the product workflow.
The implementation workflow
Read the evidence.
A prioritised finding register, not a silent repair.
Explore Portal Audit