A forecast review starts with a familiar slide: pipeline coverage looks healthy, conversion rates appear stable, and the VP of Sales is ready to defend the number. Then someone checks the underlying CRM records and finds duplicate contacts, stale lifecycle dates, inconsistent campaign values, and opportunities with missing or unreliable fields. The dashboard didn't fail on its own. The operating data behind it did.
That's why data quality metrics belong in RevOps, marketing operations, and sales operations meetings, not only in technical audits. A useful programme connects CRM conditions to pipeline accuracy, attribution, forecasting, routing, and AI readiness across Salesforce Sales Cloud, Account Engagement, Service Cloud, Revenue Cloud, and HubSpot Sales and Marketing Hubs.
Why RevOps Teams Need Data Quality Metrics
A forecast can look precise while resting on weak inputs. A missing close date changes the opportunity's time bucket. An inconsistent stage value changes conversion analysis. A duplicate account can receive multiple owners, enter more than one routing path, and inflate the addressable market used in account-based planning.
The problem is rarely one bad record. CRM decay compounds inside the calculations that leaders use to allocate budget and judge performance. A stale MQL date distorts speed-to-lead reporting. A free-text industry value prevents segmentation from matching. An unvalidated email can damage campaign execution and weaken confidence in every dashboard that depends on contactability.
Practical rule: If a data problem can change who gets routed, what gets attributed, or which opportunities appear commit-ready, it belongs on the RevOps scorecard.
The six operating levers are completeness, accuracy, timeliness, uniqueness, consistency, and validity. They describe whether required information exists, reflects reality, arrives when needed, appears only once, agrees across systems, and conforms to defined rules. That framing aligns with the broader public-sector view of data quality as fitness for use. Statistics Canada's quality framework treats relevance, accuracy, timeliness, accessibility, interpretability, and coherence as distinct dimensions, which is a useful reminder that a complete record can still be wrong or unsuitable for a particular decision.
California's governance programmes offer a practical precedent. Caltrans formally brands its enterprise data governance programme CTDATA, short for “Caltrans Data is Authoritative, Trusted, and Accessible,” and its CTDATA overview positions authoritative ownership and controlled publishing as operating disciplines. For B2B teams, the lesson is straightforward: CRM data shouldn't be treated as an ad hoc byproduct of sales activity. It needs owners, definitions, review paths, and measurable release standards.
AI raises the bar further. A model can process a complete record that contains the wrong industry, a stale employee count, or a contradictory lifecycle definition. Teams that want to build a data-driven startup culture need more than populated fields. They need structured records, source provenance, semantic consistency, and timestamps that reflect current operating reality.
The Six Core Data Quality Dimensions Explained
Use the following reference when you're defining Salesforce reports, HubSpot properties, warehouse tests, or SQL checks. The formula matters, but the business definition behind the denominator matters more. A completeness score that includes optional fields can make a healthy database look weak, while a score that ignores critical opportunity fields can hide forecast risk.
| Dimension | Formula | Measurement Method |
|---|---|---|
| Completeness | Populated required fields ÷ required field instances × 100 | Count non-null values for required fields on Lead, Contact, Account, and Opportunity records. |
| Accuracy | Sample-verified correct values ÷ sampled values × 100 | Compare a controlled sample with an authoritative source, enrichment record, or approved business record. |
| Timeliness | Median age of key lifecycle timestamps, or records within SLA ÷ applicable records × 100 | Calculate the age of MQL_Date, Stage_Changed_Date, Last_Activity_Date, or CloseDate against the reporting cut. |
| Uniqueness | Duplicate records ÷ total records × 100 | Group by normalised email, company domain, or a composite key such as domain plus country. |
| Consistency | Matching cross-system or cross-object values ÷ comparable values × 100 | Compare shared fields, such as Account.BillingCountry and Contact.MailingCountry, after applying the same normalisation rules. |
| Validity | Values passing format and rule checks ÷ tested values × 100 | Test email patterns, phone formats, country codes, picklist membership, dates, ranges, and required business logic. |
Completeness and accuracy
Completeness answers, “Is a value present?” A Salesforce-style query can identify missing opportunity amounts:
SELECT Id FROM Opportunity WHERE Amount = NULL AND IsClosed = FALSE
For HubSpot, create a filtered view or active list where required contact or company properties are unknown. Measure both record-level completion and field-level completion. A record may contain an email but still lack the account, industry, or lifecycle timestamp required by the workflow.
Accuracy asks whether the value is correct. It normally needs sampling because the CRM doesn't know reality by itself. Select tiered records, compare company domain and employee count with an approved source, and record pass or fail outcomes. Statistics Canada's guidance is useful here because it distinguishes accuracy from completeness and describes accuracy in terms of bias and variance. A field can be populated consistently and still be systematically wrong.
Timeliness and uniqueness
Timeliness measures whether data arrives while it can still support the decision. For a sales process, calculate the median age of key timestamps and separately track records outside the agreed service level. California's CalOMS reporting framework demonstrates why this needs recurring operational monitoring. Its measures include late submissions, late resubmissions, providers not reporting, and rejected records in a selected reporting month, as shown in the CalOMS data-quality training report.
Uniqueness prevents one real-world entity from becoming several analytical entities. Normalise case, whitespace, punctuation, and common domain variations before grouping. In Salesforce, Contacts should have zero tolerance for duplicate email values where email is the identity key. For Accounts, combine domain, legal name, country, and parent relationship when a domain alone isn't sufficient.
Consistency and validity
Consistency checks agreement. Compare Account and Contact country fields, lifecycle stage across marketing automation and CRM, and campaign naming between HubSpot and Salesforce. Validity checks whether a value conforms to a rule, not whether it describes reality. A syntactically valid email may still be undeliverable, while a country value can be spelled correctly but fail the approved ISO-code list.
Keep the dimensions separate in your dashboard. Combining them into one “health score” makes it difficult to decide whether the fix belongs in form validation, enrichment, integration mapping, ownership, or process training.
Thresholds and Benchmarks That Actually Work
A threshold is useful only when someone has agreed to act after a breach. Avoid the usual “target 100%” advice. Some fields are expensive to enrich, some records legitimately lack a value, and forcing teams to complete irrelevant fields creates fake data that looks tidy but harms trust.
For a mature B2B funnel, use these starting commitments:
| Dimension | Threshold | Business Impact |
|---|---|---|
| Completeness | 95% or higher for required GTM fields, 99% for email, account, and opportunity amount | Protects routing, segmentation, pipeline valuation, and forecast inclusion. |
| Accuracy | 90% sample-verified for tier-one accounts, 85% for tier-two accounts | Keeps strategic account planning and enrichment decisions grounded in reality. |
| Timeliness | Lifecycle stage updated within 7 days, last activity within 14 days | Reduces stale funnel views and prevents aged activity from masking stalled deals. |
| Uniqueness | Duplicate rate below 2%, with zero tolerance for duplicate Contact emails | Prevents inflated TAM, conflicting ownership, and duplicated campaign or sales activity. |
| Consistency | Cross-object field agreement above 98% | Supports reliable attribution, routing, and reporting across Salesforce and HubSpot. |
| Validity | Above 99% for email, phone, and country formats | Stops avoidable integration failures and reduces unusable audience records. |
These are operational recommendations, not universal laws. A strategic account list may justify stricter accuracy requirements than a broad prospecting segment. Enrichment fields can tolerate a lower threshold when the field isn't used for routing, scoring, or forecast logic. Critical identity and financial fields shouldn't receive the same tolerance.
The business cost also varies by funnel maturity. A five-point completeness drop can correlate with roughly 7% to 10% pipeline leakage in mature funnels, but that relationship depends on field criticality, workflow design, and how the team defines leakage. Treat it as a sensitivity signal, not a promise. MarTech Do's data quality assurance guide is useful when translating these dimensions into a KPI framework across accuracy, completeness, consistency, timeliness, coherence, interpretability, relevance, and reliability.
Thresholds need an owner and an exception path
Assign one accountable owner to every threshold. Marketing operations can own campaign and lifecycle fields, sales operations can own opportunity and activity data, and a RevOps leader can own cross-system definitions and escalation.
Review thresholds quarterly. California's public reporting systems show how mature programmes make cadence explicit. The California Air Resources Board annual data quality report measures capture performance with defined reporting outcomes, while California workers' compensation reporting distinguishes accepted, accepted with errors, and rejected records in its acknowledgment quality table. RevOps teams can apply the same principle by separating immediately usable records from records requiring remediation.
Dashboards and Alerting Patterns for CRM Data
A dashboard becomes operational when every panel answers three questions: what failed, who owns it, and when must it be fixed. Without those fields, a scorecard becomes wallpaper for a weekly meeting.
Build three layers rather than forcing one platform to do everything.
Salesforce-native layer. Use custom report types for Accounts, Contacts, Leads, and Opportunities. Formula fields can flag missing critical values, invalid country codes, old lifecycle timestamps, and suspicious email patterns. Report folders should separate daily exception views from trend reporting. A sales operations owner can manage opportunity completeness, while marketing operations owns Lead and Contact validity.
HubSpot operations layer. Create custom contact and company properties for validation status, enrichment source, last quality check, and remediation status. Active lists can isolate unknown required properties, duplicate candidates, and records outside freshness rules. Keep the operational lists close to the teams who can correct the data, rather than sending every issue to a central analyst.
GTM engineering layer. Replicate CRM and marketing data into Snowflake, BigQuery, or Postgres, then surface reconciled tests through Looker, Mode, or Hex. This layer should compare systems, not merely reproduce them. It's where you detect a lifecycle stage that differs between HubSpot and Salesforce, or an Account whose owner and territory disagree with routing logic.
| Metric | Cadence | Surface | Alert Channel | Owner |
|---|---|---|---|---|
| Inbound email and field validity | Daily | Salesforce reports, HubSpot active lists | Slack webhook | Marketing operations |
| Duplicate Contacts and Accounts | Daily | CRM exception view and warehouse test | Slack webhook | RevOps engineering |
| Required-field completeness on new Accounts and Opportunities | Weekly | Native dashboards and warehouse trend | Weekly digest email | Sales operations |
| Cross-system consistency | Weekly | Looker, Mode, or Hex | Slack and ticket | RevOps systems owner |
| Accuracy against enrichment baselines | Monthly | Warehouse trend dashboard | Leadership digest | Data steward |
| Forecast-field reconciliation | Monthly and before forecast calls | Salesforce forecast dashboard | Escalation email | Sales leadership |
Use alerts for action, not noise
A serverless function can evaluate test results and post a severity-tagged Slack message when a threshold breaks. Include the failing metric, affected object, record count, owner, list view or query link, and remediation SLA. Don't alert on every individual bad record unless the record is revenue-critical. Aggregate routine exceptions and reserve urgent alerts for routing, identity, or forecast fields.
Send leadership a weekly digest with current values, week-over-week deltas, unresolved tickets, and ageing. The unified RevOps dashboard architecture for HubSpot and Salesforce provides a useful reference for designing a cross-platform view without hiding system-specific ownership.
Remediation Workflows and Governance Routines
A quality breach should create work, not merely a red tile. A practical loop starts when an alert lands in a severity-tagged Slack channel with a Salesforce list-view link or HubSpot active-list link. The owner triages the issue by affected object, business impact, source, and remediation deadline.
Clay can serve as the enrichment and deduplication layer through Clay. The workflow should remain controlled, especially when data flows back into Salesforce or HubSpot.
The five-stage remediation loop
Triage: The RevOps owner confirms whether the breach is a data-entry issue, an integration failure, a definition conflict, or an enrichment gap. Severity should reflect the downstream consequence, not the raw record count.
Dedupe: Clay normalises email addresses and domains, then compares suspected duplicates against Clearbit and Apollo data. It should produce a candidate list with match reasoning, not automatically merge records.
Enrichment: Approved firmographics, industry values, and employee counts can flow back to HubSpot and Salesforce through a webhook. Preserve source and checked-at fields so a later reviewer can distinguish enrichment from seller-entered information.
Validation: A RevOps engineer runs regression checks before merging or updating records. Test routing, lifecycle automation, campaign association, territory assignment, and reporting joins. A technically clean record can still break a downstream process if its value changes the meaning of a workflow.
Prevention: Add form validation, controlled picklists, lead-routing rules, duplicate checks, and field-level permissions at the point of entry. Cleansing without prevention creates a recurring queue.
The governance routine should be light enough to survive busy quarters. Hold a weekly 30-minute data standup for open breaches and ageing tickets. Run a monthly council review with sales and marketing leads to approve definition changes. Store the quarterly audit and decision history in a Notion runbook.
Ownership rule: Only designated RevOps or data stewards should merge records. Marketing operations should manage campaign and lifecycle picklists, sales operations should manage opportunity process fields, and the scorecard owner should approve threshold changes.
Document the RACI, vendor sources, merge authority, escalation path, and remediation SLA in the runbook. A CRM data hygiene operating guide can help teams turn one-time cleanup into a repeatable operating routine.
Mapping Metrics to Pipeline, Attribution, and Forecasting
Data quality earns executive attention when the team connects each metric to a GTM artifact. Missing opportunity amount and close date fields weaken forecast coverage. Duplicate Accounts inflate the denominator used for TAM and can send the same buying group through multiple routing paths. Stale firmographics distort ICP scoring and account prioritisation.
Attribution has its own failure mode. If HubSpot and Salesforce use different UTM or campaign-name values, the systems can't reliably reconcile source, influence, and pipeline credit. The issue may not appear as a broken integration. It appears as a marketing-sourced number that changes when someone groups values differently.
| Quality Metric | GTM Artifact Affected | Business Symptom |
|---|---|---|
| Completeness | Opportunity forecast and lead routing | Deals disappear from forecast views or enquiries wait for ownership. |
| Accuracy | ICP scoring and account planning | Reps prioritise companies that don't match the intended market. |
| Timeliness | Stage conversion and activity reporting | Leaders read old funnel conditions as current performance. |
| Uniqueness | TAM, territory, and pipeline coverage | The same company appears multiple times and inflates coverage. |
| Consistency | Campaign attribution and lifecycle reporting | Marketing and sales reports assign different meanings to the same record. |
| Validity | Integrations, segmentation, and deliverability | Automations fail or audiences include unusable records. |
Sensitivity analysis makes the discussion concrete, but it must be labelled as a scenario rather than presented as a universal benchmark. A 5% miss in opportunity close dates can shift a quarterly forecast by 8% to 12% in a model where close dates drive period allocation. A 10% duplicate Account rate can inflate pipeline coverage ratios, while inconsistent campaign names can reallocate 20% of marketing-sourced credit in an attribution model. These figures are useful stress tests only when your team applies them to its own funnel definitions and validates the result.
Assign every metric a named owner and remediation SLA. Put the scorecard in QBR materials, reconcile quality exceptions before forecast calls close, and require an explicit data sign-off for material changes. That turns data quality from an IT hygiene project into a forecasting control.
Your 30/60/90 Data Quality Rollout Plan
Start with measurement, not a cleansing marathon. A phased rollout gives the team a baseline, creates automation, and then embeds governance before the programme depends on one person's memory.
Days 0 to 30
Stand up the six metrics in one dashboard across Salesforce, HubSpot, or the warehouse layer. Baseline each metric by object, segment, owner, and source. Select two high-impact fields first, typically company size and email deliverability, then trace each failure to its point of entry.
Days 31 to 60
Deploy Salesforce validation rules, restricted picklists, required-field logic, and duplicate management. In HubSpot, use property validation, active lists, workflows, and controlled lifecycle definitions. Connect Clay enrichment to lead routing, then launch weekly Slack alerts for completeness and validity breaches.
Days 61 to 90
Institutionalise the programme with a documented RACI, monthly metric reviews, quarterly threshold calibration, and a feedback loop into campaign and lifecycle operations. Add forecast reconciliation to the review so the team can show whether quality exceptions are changing pipeline interpretation.
Use this owner-assignable checklist:
- Metric owner: Name the accountable person for each dimension and object.
- Alert threshold: Record the trigger, severity, notification channel, and response SLA.
- Dedupe cadence: Define matching keys, review authority, and merge procedure.
- Enrichment vendor: Document Clay, Clearbit, Apollo, or another approved source and its provenance fields.
- Forecast reconciliation: Specify who checks opportunity completeness and timing before forecast sign-off.
- Change control: Store field definitions, picklist changes, mappings, and exceptions in the runbook.
Completeness alone doesn't make a CRM AI-ready. Semantic consistency, source provenance, and timeliness determine whether models trained on Salesforce or HubSpot records can generalise beyond the clean sample your team happened to review.
MarTech Do helps B2B teams audit and improve Salesforce and HubSpot environments, including CRM data remediation, deduplication, lifecycle governance, integrations, and RevOps dashboards. If your data quality metrics aren't yet connected to routing, attribution, or forecasting, visit MarTech Do to discuss a practical audit or implementation plan.
