What a Working HubSpot Setup Actually Looks Like (Audit Checklist)

If your sales team is exporting HubSpot data to Excel every Friday, your setup is broken. The export is a confession — somebody on the team has stopped trusting the dashboard and is rebuilding it by hand. By the time that habit is two months old, the portal has become a glorified storage layer for an Excel-based revenue operation. That is the most common shape of a broken HubSpot, and it is rarely diagnosed until something hits a board deck.

This is the 50-point audit we run on every new client portal before we touch a single workflow. It is the same audit our senior operators run on every NEOME-built implementation before handover. The goal is not to find every flaw — every portal has flaws. The goal is to figure out, in a half-day, whether what you have is healthy, degrading, struggling, or broken, and then to know what to fix first.

Most portals we audit fail 30+ of these checks. That number is not a sales line; it is the single most consistent finding across 50+ audits. HubSpot is hard to keep clean. The good news is that going from “30 fails” to “45+ passes” is usually 8–12 weeks of focused remediation, not a fresh 6-month implementation.

You can run this on your own portal in about 90 minutes. Read the section at the bottom on how to do that.

The 10-minute gut-check (3 questions)

Before the 50-point audit, three questions tell you 80% of the answer.

1. Is your sales team exporting HubSpot data to Excel weekly? If yes, your reporting is broken. The sales team has lost confidence in the dashboards and is rebuilding numbers by hand. This is the single strongest signal of an unhealthy portal — stronger than any individual technical metric.

2. Can your CMO and your VP Sales agree on the number of MQLs last quarter? If they cannot, your lifecycle definitions are broken. The two halves of the funnel are operating on different scoreboards. This becomes a board-level problem fast — usually the next time someone asks “why isn’t pipeline converting?”

3. When was the last time anyone audited a workflow that stopped firing? “We don’t have a process for that” is the answer most teams give. Workflows fail silently in HubSpot — no alarm, no email, no flag. If nothing watches them, roughly 10% are quietly dead at any given time. The longer the portal runs, the higher that number climbs.

If you answered “yes / no / we don’t” to those three, the rest of the audit will confirm what you already suspect. Run it anyway — knowing which layers are broken is what determines whether you patch or rebuild.

Layer 1: Data layer audit (12 checks)

The data layer is the foundation. If this is broken, nothing on top of it is reliable. Every dashboard, every workflow, every forecast inherits the rot.

Contact deduplication. Run HubSpot’s duplicate management tool. Healthy = under 2% of total contacts flagged. Anything over 5% is structural — usually multiple unsupervised import paths.
Company deduplication. Same standard. The most common breakage point is companies created from imports without domain matching enabled.
Email validity. What share of contacts have a valid, deliverable email? Healthy = 90%+. Below 80% means inbound and import hygiene have failed and your sender reputation is taking damage.
Phone number formatting. Are phone numbers stored consistently (E.164 ideally)? Mixed formats break SMS, dialers, and most third-party dialer integrations.
Lifecycle stage definitions documented. A one-pager defining what “MQL” means in your business, with criteria. If this does not exist, lifecycle stages are vibes and your funnel report is fiction.
Lifecycle stage progression rules. Do contacts move through lifecycle stages via workflow, manually, or both? Healthy = workflow-enforced with documented exceptions. Mixed mode without documentation is the most common source of duplicate-MQL arguments.
Required-property fill rate. What share of “required” properties are actually filled? Anything below 80% on a “required” property means the requirement is not being enforced.
Custom property sprawl. Total custom properties per object. Over 200 is usually a sign of accumulated mess — old experiments that nobody deleted.
Property naming convention. Properties should follow a documented convention (snake_case, prefix by Hub, source-tagged). Inconsistent naming is the tell of multiple uncoordinated implementers.
GDPR / consent fields. If you sell into the EU/UK, are subscription types defined and the legal basis property populated on every contact? “We meant to do that” is the answer that gets you fined.
Source-of-truth for the Account record. Who owns Account-level data — sales, marketing, ops? “Whoever updated it last” is a failure mode dressed up as flexibility.
Data sync direction documented. For every integrated system (Salesforce, Gong, ZoomInfo, Clearbit, billing), is sync direction written down? Undocumented sync is a guaranteed future failure.

Layer 2: Marketing Hub audit (10 checks)

Workflow inventory. How many active workflows? Healthy portals usually have 30–80. Over 200 active workflows is a red flag — most teams cannot mentally inventory that many, and what cannot be inventoried cannot be maintained.
Workflow ownership. Does every workflow have a documented owner? Orphaned workflows are landmines that go off when a team member leaves.
Lead scoring model documented. A one-pager explaining the score thresholds and what each tier triggers. “We have lead scoring” without a document is not a model; it is a black box.
Score recalibration cadence. When was scoring last recalibrated against closed-won data? Anything older than 12 months is stale. The market your scoring was built for is not the market you are selling into now.
Email deliverability. Inbox-placement rate for the last 30 days. Below 90% is a problem; below 85% means you are training Gmail to bury you.
Subscription management. At least 3 subscription types and a working preference center. Single-list-everyone setups age into the kind of unsubscribe rate that kills a domain.
Form-to-CRM property mapping. All form fields mapped to the right CRM properties. The most common silent break: hidden fields not syncing because the form was duplicated and the mapping was not.
Attribution model. One configured model, with a documented rationale. “All models on, look at whichever” is the failure pattern that lets every team claim credit.
Campaign tracking. Campaign records created for every paid channel and every major content push. Without campaign records, attribution is partial fiction.
MQL → SQL handoff workflow. Automated handoff with a documented SLA (usually 5–15 minutes for inbound demo requests). Manual handoffs leak leads on weekends and holidays.

Layer 3: Sales Hub audit (10 checks)

Pipeline stage definitions. Each stage has a written entry criterion and a written exit criterion. “Discovery” with no definition means every rep defines it differently and the pipeline number is a guess.
Stage probability calibration. Probability per stage matches historical close rate within 10%. Most setups have stages locked at default probabilities (10/25/50/75/90) that no longer reflect reality.
Required deal properties at each stage. Stage gating prevents incomplete records. Without it, pipeline data is missing exactly where it matters most — at the late stages where forecast lives.
Forecast accuracy. Variance between forecasted and actual closed pipeline last quarter. Healthy = within 15%. Anything over 25% means forecast cannot be used for hiring or planning.
Activity logging. Average activities per closed-won deal. If deals close with three logged activities, sales is not using HubSpot — they are using something else and updating HubSpot as theatre.
Deal-to-Company association. Every deal associates cleanly to a Company record. Orphan deals are usually a billing or import issue and they distort account-level reports.
Sales rep ownership. Every active deal has an owner who is still at the company. Stale owners on active deals indicate the offboarding process never reaches HubSpot.
Quote workflow (if used). Approvals routed and tracked. The most common gap: quotes generated outside HubSpot (in Google Docs) and pasted in as a PDF, breaking every CPQ-style report.
Sequences inventory. Active sequences match active sales motions. Sequences from a discontinued ICP are still firing in roughly half the portals we audit.
Sales reporting trust. Ask three reps independently: “Is the pipeline number on the dashboard correct?” Three “no”s means the dashboard is theatre and they are running a shadow spreadsheet.

Layer 4: Service / CS audit (8 checks)

Ticket categorisation. Tickets categorised against a defined taxonomy. Free-text “subject” with no taxonomy is unfit for trend analysis and unfit for product feedback.
SLA tracking. First-response and resolution SLAs defined and tracked. Missing SLAs mean you cannot tell whether support is healthy or quietly drowning.
Knowledge base coverage. Top 20 ticket categories — does each have a public KB article? Below 70% coverage means you are paying humans to answer the same question repeatedly.
Health score property. Customer health score that updates automatically from product, support, and engagement signals. “We score health in a spreadsheet” means CS is operating off-system.
Renewal pipeline. A separate Renewal pipeline (or stage set). Renewals living in the new-business pipeline distort forecast and bury at-risk accounts.
Expansion-revenue tracking. Upsell and cross-sell deals identifiable in the data via deal type or pipeline. “Tagged in the deal name” is not a real implementation.
NPS / CSAT integration. Survey scores syncing to Contact and Company records. Survey data living only in the survey tool is a missed integration and an invisible early-warning system.
Onboarding playbook automation. New-customer onboarding running as a workflow with checkpoints. Manual onboarding has the highest CS-team load and the most variability per customer.

Layer 5: Reporting audit (10 checks)

Dashboard ownership. Every dashboard has an owner. “Whoever built it” almost always means nobody, and nobody-owned dashboards are nobody-trusted dashboards.
Dashboard count. 5–15 dashboards is healthy. Over 50 means dashboard sprawl — too many sources of conflicting truth.
Saved-report inventory. Same logic as dashboards. Over 200 saved reports is a smell; over 500 means nobody can find anything.
Funnel report integrity. Marketing funnel report agrees with the sales pipeline report on the MQL → SQL number. Disagreement is a definition problem and it is the most common cause of CMO/VP-Sales QBR fights.
Attribution report sanity. Attribution dashboards add up to total revenue. First-touch and last-touch should both equal total revenue with different channel mixes. If totals disagree, a property is broken upstream.
Custom report performance. No reports timing out. Report timeouts almost always indicate a bad property structure or a missing index — they are a sign of underlying data debt.
Date-range hygiene. Reports use explicit, consistent date ranges. “Last 30 days” reports compared to “this month” reports create the kind of executive confusion that kills credibility.
Exec dashboard freshness. Exec dashboard updates weekly with no manual intervention. “Manually refreshed before the QBR” is a failure mode pretending to be a process.
Forecasting reports vs Sales Hub forecast. Two HubSpot views of the same number agree. They often do not — usually because one is filtering closed-lost out and the other is not.
Reporting access permissioned by role. Everyone-sees-everything dashboards are an invitation to render bad numbers public, and they leak compensation and pipeline information that should be scoped.

How to score it

Tally your “yes, we have this and it works” answers.

45+ Healthy. The implementation was good or you have a strong RevOps function maintaining it. Optimise at the margins.
35–44 Working but degrading. The portal will hold for now, but failure modes are accumulating. Plan a focused remediation in the next quarter before the gut-check questions start coming back as “yes.”
25–34 Struggling. Several layers are not load-bearing. The team is already feeling it; you may not have heard about it yet because nobody wants to be the one to flag it. This is the most common state of HubSpot portals 18–36 months after implementation.
Below 25 Broken. Patches will not fix this — the foundation needs to be redone. The sooner the rebuild starts, the less data damage you carry forward.

The “what to fix first” prioritisation framework

A failing audit is overwhelming. The temptation is to fix everything; the right move is to fix in this order:

1. Data layer first, always. Anything in Layer 1 that fails is upstream of every other failure. Dedup, lifecycle definitions, and required-property enforcement come before workflow cleanup. There is no point fixing a broken workflow that runs on broken data.

2. The fail your team already knows about. Whatever the sales team complains about loudest in standups — pipeline accuracy, lead routing, forecast — fix that visibly and fast. Adoption is the close. People remember when the portal got better in the way they noticed.

3. The silent failures that are bleeding pipeline. Dead workflows, unmapped form fields, broken handoffs. These are not loud, but they are the most expensive — they cost you deals that nobody knows you lost.

4. Reporting last. Reporting only matters once the data and process feeding it are correct. Rebuilding dashboards before the underlying data is fixed just gives you prettier wrong numbers.

This order is non-negotiable in our practice. We have watched teams spend three weeks rebuilding dashboards on top of a broken data layer and end up with the same QBR fights they started with.

How to use this checklist on your own portal in 90 minutes

You do not need to hire anyone to run this. You need 90 minutes, admin access, and a willingness to mark things as broken even when you built them.

Minutes 0–10. Run the three gut-check questions. Write the answers down before you start digging.
Minutes 10–35. Layer 1 (data). The duplicate management tool, the lifecycle definitions doc, the property fill rates. This is the slowest layer and the most important.
Minutes 35–55. Layers 2 and 3 (Marketing and Sales). Most of these you can answer from the workflow list, the pipeline view, and a quick conversation with whoever owns each.
Minutes 55–70. Layer 4 (Service). If you do not have Service Hub, skip; if you do, walk the ticket list and ask CS leadership the SLA question directly.
Minutes 70–85. Layer 5 (Reporting). Open every dashboard. Ask “who owns this and do we trust it.” If the answer is “nobody / no” mark it failed.
Minutes 85–90. Tally. Score yourself honestly. The portal does not get better because you scored kindly.

If the score lands below 35, the next step is not to start fixing. The next step is to figure out whether you are looking at a remediation project (12 weeks of cleanup) or a rebuild. We wrote up that decision separately because it is the single most consequential call most teams get wrong — they replatform when they should remediate, or they patch when they should rebuild.

If you want a second pair of eyes on the audit, that is what we do. We run the same 50-point checklist on every prospect’s portal as a 4-hour paid engagement, and the deliverable is the marked-up version plus a prioritised remediation plan. We also run it ourselves on every NEOME-built implementation before handover — the agent ships the configuration and the senior operator runs the audit. (More on that workflow in Inside NEOME.)

What to do next

Run the checklist on your own portal. If you score 45+, leave it alone and revisit in a year. If you score 35–44, plan a focused remediation. If you score below 35, the next read is How to Fix a Broken HubSpot Portal Without Replatforming — most “we need a new portal” instincts at this score are wrong.

If you would rather we run the audit, book a free 30-min consultation with your portal hub ID and the three gut-check answers. We will come back with a marked-up checklist and a 30/60/90 plan inside a week.