Most HubSpot audits produce a list of settings that differ from best practice. That list is not useful on its own, because best practice for a 200-person enterprise is wrong for a 30-person startup. A useful audit measures the portal against your actual revenue motion, and ranks every finding by what it costs you.
This is the checklist we run in our own HubSpot audits. You can run it yourself.
Why does the order of the audit matter?
Because problems propagate downward. A dashboard showing wrong numbers is rarely a reporting problem. Trace it back and you usually find a lifecycle stage that half the database never received, or two properties holding different versions of the same fact. Fix the chart without fixing the stage, and you have moved the wrong number somewhere prettier.
So the audit runs in the order that problems flow: data model first, reporting last. Findings in layer one explain most of what you will see in layers four and five.
Layer 1: What should you check in the data model?
The data model is the set of objects, properties, and associations everything else is built on. Check four things:
- Duplicate properties. Two custom properties holding the same fact, filled by different teams. Every one of these is a future reporting argument.
- Duplicate records. Companies and contacts that exist twice. In one engagement we imported and deduplicated more than 50,000 companies in a single batch, for a portal where duplicates were arriving faster than anyone cleaned them. The fix started at the point of entry, not with a cleanup script.
- Unused properties. Properties with no values, or values on under 5% of records. They are noise that makes every picker and report harder to use.
- Association logic. Whether contacts, companies, and deals are actually linked the way your reporting assumes they are.
Layer 2: Are lifecycle stages and lead status actually defined?
Almost every portal has lifecycle stages switched on. Very few have definitions that marketing and sales both agreed to. Check:
- What share of contacts have no lifecycle stage at all. In portals we audit, 30 to 60% is common, and it means funnel reporting is fiction.
- Whether anyone can say, in one sentence, what makes a contact an MQL. If the answer starts with “it depends,” you found the problem.
- Whether stages only move forward. Contacts silently sliding backward is one of HubSpot’s classic quiet failures.
HubSpot’s own lifecycle stage documentation covers the mechanics. The audit question is not whether the feature works. It is whether your team shares one definition per stage.
Layer 3: Can your scoring and routing drift?
Scoring models degrade in a specific, checkable way. HubSpot recalculates scores as new actions arrive, and later low-value actions can overwrite a higher score, quietly demoting a lead that already qualified. We found this the hard way and now anchor scoring models with static lists, so a qualified contact stays qualified.
For routing, check the unowned-contact count. In the portals we see, 20 to 40% of contacts have no owner. Every unowned inbound lead is a lead nobody is going to call.
Layer 4: Which automated emails are colliding?
Nobody has the whole picture of automated email, because nobody owns all of it. Marketing runs some sequences, sales runs others, and customer success added a few during onboarding. Individually each one is reasonable. Together they can send the same contact four emails in one week.
The check: map every automated email onto a single 30-day timeline per contact type. Conflicts and overlaps become visible immediately. We productized this as a customer journey email audit because it kept finding collisions no one knew about.
Also list workflows that have not fired in 90 days, and workflows with overlapping enrollment criteria. Both are common, and both cause the “why did this contact get that email” tickets.
Layer 5: Which reports have one agreed definition?
The final layer is fast to check and slow to fix. For each number leadership looks at weekly, ask: is there exactly one definition, in writing, that every team accepts? Pipeline, MQLs, win rate, ARR. If sales and finance calculate ARR differently, the audit finding is not “fix the report.” It is “get one definition signed off, then fix the report.”
One measurable outcome from this layer in our own work: a client’s seat allocation had drifted so far from actual usage that right-sizing it to 28 seats avoided an Enterprise upgrade they were about to pay for.
What should the audit deliver at the end?
A findings document ordered by cost, not by severity label. Each finding should state what is wrong, what it costs in pipeline visibility or rep time, what fixing it involves, and roughly how long that takes. A finding that cannot name its cost does not belong in the report.
Then decide who fixes it. A well-written findings document can go straight to an internal team. If you want outside help instead, our HubSpot audit service runs exactly this process in about two weeks.