A reporting dispute often looks like an analytics problem. One dashboard says 1,240 customers, another says 1,198, and a spreadsheet says 1,311. The instinct is to fix the report. But all three outputs may be faithfully representing three different definitions or data paths.
PwC's 2026 operations research found that poor data quality remains a major obstacle to creating value from digital initiatives. The practical lesson is that reporting trust is built upstream.
If two reports disagree, do not start by asking which chart is wrong. Ask what each system believes the number means.
Five common causes
Different definitions
“Customer”, “active account”, “revenue”, “qualified lead” and “completed order” can mean different things across departments. A metric without a controlled definition is an argument waiting to happen.
Different systems of record
Sales may trust the CRM while finance trusts the accounting platform. If ownership of the underlying data is unclear, both can be locally correct and operationally incompatible.
Different timing
One report may be real-time, another refreshed nightly and another manually exported on Friday. Apparent data-quality problems can actually be synchronisation problems.
Manual reconciliation
When people copy, clean and reshape data in spreadsheets, undocumented logic enters the reporting chain. That may be necessary temporarily, but it creates another transformation layer that needs ownership.
Bad capture at source
A dashboard cannot repair information that was never captured, captured inconsistently or stored in free text when structured data was required.
Trace the number backwards
Reliable reporting begins with lineage. For an important metric, identify its definition, source system, source fields, transformation rules, refresh timing, exclusions and owner. Then trace the number from the report back to the operational event that created it.
This exercise frequently reveals that the organisation does not have a dashboard problem. It has an operating-definition problem or a systems-architecture problem.
Do not create a “single source of truth” by slogan
Centralising data can help, but declaring one database the source of truth does not settle which business process owns each fact. Customer identity may originate in a CRM, invoice status in finance and fulfilment status in an operations platform. The architecture needs explicit authority by data domain.
That distinction becomes more important as AI consumes operational information. McKinsey's 2026 work on agentic AI emphasises governed, reusable data foundations because agents need information they can interpret and trust. If humans cannot reconcile a metric, an autonomous workflow inherits the ambiguity.
What better looks like
Trusted reporting has controlled metric definitions, explicit systems of record, known refresh timing, visible transformation logic and named ownership. Exceptions are explainable rather than mysterious. Manual reconciliation is reduced or at least documented.
The goal is not perfect data everywhere. It is sufficient reliability around the information used to operate and make decisions.
When to act
Repeated meetings to reconcile numbers, spreadsheet adjustments before every management report, departments maintaining private definitions and executives hesitating to trust dashboards are signs that reporting has become an operational control issue.