Data Quality for Reporting: A Single Source of Truth

A report is only as trustworthy as the data behind it. When two reports quote different numbers for the same thing, when a figure cannot be traced to its source, or when last month’s number changes this month, the report loses its authority – and the reader starts verifying everything, which is the beginning of the end for the reporting function.

This guide is the hub for data quality in reporting. It covers what a single source of truth really means, the dimensions of data quality, the checks to run before every report, reconciliation and lineage, and what to do when a figure is wrong. It is the trust layer beneath every report you produce.

Reporting is a promise that the number means what it says. Data quality is how you keep that promise.

Why data quality is the real reporting risk

Most reporting pain is blamed on production – the deadline, the formatting, the chasing. The deeper problem is usually data. When inputs are fragmented and definitions are inconsistent, production becomes a reconciliation exercise every cycle, and the report that emerges is only as reliable as the weakest source behind it.

The documented cost is large. Gartner’s benchmark put the average annual cost of poor data quality at about $12.9 million per organization, based on a survey of 154 enterprise customers. IBM’s Institute for Business Value reported that 43% of chief operating officers rank data quality as their most significant data priority, and that over a quarter of organizations estimate losses of more than $5 million a year from poor data quality. A Harvard Business Review study found that 47% of newly created data records contain at least one critical error, and only 3% of companies’ data met basic quality standards.

In reports, that risk surfaces as the two failures readers never forgive: numbers that do not match, and numbers that cannot be traced.

What a single source of truth actually means

A single source of truth (SSOT) is not one database or one dashboard. It is an agreement about where the authoritative version of each metric lives, and a discipline that every report draws from it rather than from a local copy.

In practice, an SSOT means:

• One authoritative definition per metric, documented and shared.
• One authoritative source per metric, named and owned.
• One refresh cadence per metric, understood by everyone who reports it.
• One place a reader can go to resolve a disagreement.

Most firms do not need a data platform to get most of the benefit. They need a metric dictionary, named owners and a rule that reports are generated from the agreed source – not from a spreadsheet someone maintained separately. See What Is a Single Source of Truth?.

The dimensions of reporting data quality

Data quality is not one property. Six dimensions matter for reports, and each can fail independently.

A report can be accurate and still fail on consistency, if the same metric is defined differently in two places. That is why the dictionary matters as much as the data itself.

Data-quality checks before every report

A short, repeatable set of checks catches most report errors before they reach a reader.

• Reconcile to the prior period. If a number changed, know why.
• Reconcile to the source system. Tie every headline figure back to its origin.
• Check completeness. Confirm every expected entity or line is present, not just the ones with data.
• Test internal consistency. Totals equal their parts; percentages sum correctly.
• Check period alignment. Confirm the reporting window matches the label.
• Spot-check outliers. A figure that looks surprising may be wrong, not good news.
• Confirm definitions. The metric means what the dictionary says it means.

Run them against a checklist so they are not skipped under deadline. See Data-Quality Checks for Report Production.

Reconciling numbers across reports

The most visible failure is two reports quoting different numbers. The cause is almost always a different source, a different definition or a different period.

• Name the authoritative metric and use it everywhere.
• Align periods so “this month” means the same window in every report.
• Reconcile deliberately when two systems disagree – do not average them.
• Document any intentional difference (for example, a management versus statutory basis).

For recurring reports, reconcile against the previous cycle as well as other current reports, so a number does not silently drift. See Reconciling Numbers Across Reports.

A KPI and metric dictionary

The metric dictionary is the single most leveraged artifact in reporting quality. It records, for every metric a report contains:

• The name and any aliases in use.
• The definition and formula.
• The source and system of record.
• The owner accountable for it.
• The refresh cadence and reporting period.
• Known limitations or caveats.

With a dictionary, two people produce the same number for the same metric. Without one, they improvise, and the report inherits the inconsistency. See Building a KPI & Metric Dictionary.

Data lineage and traceability

Traceability is what lets a report defend its numbers. For each key figure, you should be able to answer: where did it come from, what transformed it, and who approved it.

• Record the source for every headline figure.
• Document transformations – adjustments, allocations, exclusions.
• Keep the snapshot used for each report cycle.
• Name the owner who stands behind the number.

Lineage is also a governance control: it turns “the system says” into “here is exactly how this number was produced.” See Data Lineage & Traceability in Reports.

Handling estimates, ranges and forecasts

Not every number in a report is a hard fact. Estimates, ranges and forecasts are legitimate – if they are labeled and handled properly.

• Label them clearly as estimates, forecasts or ranges.
• State the method and the assumptions behind them.
• Present ranges, not false precision, where uncertainty is real.
• Do not mix forecast and actual without a visible distinction.
• Revisit prior forecasts to show how they tracked.

Presenting an estimate as a measured fact is one of the fastest ways to lose a reader’s trust. See Handling Estimates, Ranges and Forecasts.

When data is wrong: correction and disclosure

Errors happen. What matters is the response.

• Correct promptly and record the correction.
• Assess the impact – which reports and decisions the error affected.
• Disclose proportionately to those who relied on it.
• Fix the cause, not just the instance, so the same error does not recur.
• Feed it back into the checklist and the dictionary.

A controlled correction builds trust; a silent fix discovered later destroys it. See When Data Is Wrong.

A worked data-quality checklist

Run this before every report cycle.

Work it top to bottom. The checklist is deliberately short: it must be runnable under deadline, or it will be skipped.

Building the metric dictionary: a worked example

A dictionary entry is short but complete.

With entries like this, two people produce the same number for the same metric, and a reader can see exactly what a metric includes and excludes. See Building a KPI & Metric Dictionary.

Reconciliation in practice

When two systems disagree, the instinct to average is wrong. A practical reconciliation:

• Identify the difference – amount, direction and likely cause.
• Trace both numbers to their sources and definitions.
• Decide the authoritative one and document why.
• Adjust or footnote the other, rather than blending them.
• Record the reconciliation so the next cycle does not re-litigate it.

Applied consistently, reconciliation turns recurring disagreements into a settled rule.

Why spreadsheet-based reporting is risky

Spreadsheets are the workhorse of reporting, and the research on spreadsheet error is sobering. Panko’s summary of field audits found errors in about 94% of spreadsheets, and a study of 25 operational spreadsheets identified 117 confirmed errors, the largest with a $100 million-plus impact. A Harvard Business Review study found 47% of newly created data records contain at least one critical error.

The implication is not to abandon spreadsheets; it is to control them: keep one authoritative version, separate inputs from calculations, document assumptions, and reconcile to source. Human-error research is consistent that people underestimate their own error rates, which is why controls – not confidence – keep report data reliable.

The cost of poor reporting data

The cost shows up in three places: rework on numbers that should have been right, delayed decisions while leaders wait on figures they do not trust, and damaged credibility once a reader finds one wrong number and starts verifying everything.

The benchmarks are large. Gartner put poor data quality at about $12.9 million a year on average per organization, based on a survey of 154 enterprise customers, and IBM’s 2025 research found over a quarter of organizations estimating losses above $5 million a year. Even a fraction of that, applied to reporting, justifies treating data quality as a control rather than an afterthought.

Who owns reporting data quality

Ownership is the difference between a data-quality intention and a data-quality outcome.

• A data owner per source system, accountable for the data it produces.
• A reporting owner accountable for the figures in the report.
• A metric owner per metric in the dictionary.

Where ownership is unclear, quality decays. Name the owners, and record them in the dictionary so accountability survives staff changes.

Data quality across the recurring cycle

In recurring reporting, data quality has a rhythm.

• Before the freeze: confirm the data is complete and the sources are current.
• At the freeze: snapshot the data and record the version.
• After the freeze: verify figures against source and reconcile to the prior cycle.
• After issue: log any correction and feed the cause back into the checklist.

Without the rhythm, quality is checked erratically and errors slip between cycles. With it, each cycle starts from a known-good baseline.

A worked reconciliation

Suppose the finance system shows revenue of $4,200,000 for the month, while the CRM shows $4,260,000.

• Identify the difference: $60,000, with the CRM higher.
• Trace both numbers: finance is invoiced revenue; the CRM is booked revenue, including a $60,000 deal signed but not yet invoiced.
• Decide the authoritative basis for this report: finance, because the report is on invoiced revenue.
• Footnote the other if the booked-versus-invoiced difference is material to the reader.
• Record the reconciliation so the next cycle applies the same rule.

The point is not which number is “right” in the abstract, but that the report uses a defined basis consistently and can explain any difference on request.

Setting a data-quality baseline

Before improving data quality, measure it. A simple baseline records, for each critical metric:

• The source and its refresh cadence.
• The current error or rework rate – how often the figure needs correction.
• The owner.
• The last incident and its cause.

A baseline turns “our data quality feels poor” into a measured starting point, and it makes improvement visible. Without it, data-quality work is invisible, and therefore unfunded.

Common mistakes

• No metric dictionary. Different people define the same metric differently.
• Reporting from local copies. A spreadsheet maintained alongside the source system drifts.
• Reconciling by averaging. When two systems disagree, averaging hides the error.
• Presenting estimates as facts. False precision that a reader will eventually catch.
• No lineage. Numbers nobody can trace back to a source.
• Silent corrections. A fixed number with no record of what changed or why.

Frequently asked questions

What is a single source of truth in reporting?

An agreement about where the authoritative version of each metric lives, with one definition, one source, one refresh cadence and one owner – so every report draws from the same place.

Why is data quality important for reports?

Because trust is the report’s whole value. Numbers that do not match, or cannot be traced, make readers verify everything, which erodes the reporting function. Poor data quality also carries real cost – Gartner put it at about $12.9 million a year on average per organization.

What are the dimensions of data quality?

Accuracy, completeness, timeliness, consistency, validity and uniqueness. A report can be accurate yet still fail on consistency if the same metric is defined differently in two places.

How do you reconcile numbers across reports?

Name the authoritative metric, align periods, reconcile deliberately when systems disagree rather than averaging, and document any intentional difference such as management versus statutory basis.

What is a KPI or metric dictionary?

A record of every metric a report contains: its name, definition, formula, source, owner, refresh cadence and limitations. It is what makes two people produce the same number.

What should you do if a report contains a wrong figure?

Correct it promptly, record the correction and its impact, disclose proportionately to those who relied on it, and fix the underlying cause so it does not recur.

Do you need a data platform to have a single source of truth?

No. Most firms benefit first from a metric dictionary, named owners and a rule that reports draw from the agreed source. A platform helps at scale, but the discipline is what creates the single source of truth.

How do you stop a metric being defined differently by two teams?

Write it down. A metric dictionary entry – name, definition, formula, source, owner, caveats – settles disputes before they happen. In practice, most “data disputes” are definition disputes in disguise.

How often should reporting data be reconciled?

Every cycle, against both the source system and the prior period. Reconciling only to source misses drift between cycles; reconciling only to the prior cycle misses errors inherited from a faulty source.

Is it acceptable to report estimated figures?

Yes, if they are labelled as estimates, the method is stated, and they are corrected when actuals arrive. Presenting an estimate as a measured fact is what damages trust.

What is data lineage?

The record of where a figure came from, how it was transformed, and who stands behind it. It is what lets a report defend its numbers under scrutiny, and it is a governance control as much as a technical one.

How do you check a suspicious figure?

Compare it to the prior period and to the source system, confirm it against the dictionary definition, and investigate if it remains surprising. A figure that looks good but cannot be explained should be treated as suspect until it is.

How do you prioritize data-quality fixes?

By impact and frequency. Fix the metrics that appear in the most reports and cause the most rework first, and the sources that feed them. A small number of metrics usually drives most of the reporting pain.

What is a KPI dictionary used for?

To make sure two people produce the same number for the same metric. It records the definition, formula, source, owner and caveats, and it is the quiet control behind consistent reporting.

How do you handle a metric two teams define differently?

Agree one definition, write it in the dictionary, and name the owner. Where both definitions are legitimately needed – management versus statutory basis, for example – label them distinctly rather than blending them.

How do you stop data quality decaying over time?

Assign ownership, review the dictionary and the sources on a cadence, monitor error and rework rates, and feed every incident back into the checklist. Quality is maintained, not achieved once.

Is one source of truth realistic for a small business?

Yes, and it is usually simpler than for a large one. For a small firm, a single agreed spreadsheet or system, a short metric dictionary and named owners deliver most of the benefit without any platform investment.

Next step

Trust is the product of a report. Build a metric dictionary, name your sources, and run data-quality checks before every cycle. Download the Data-Quality Checklist to start, and see How to Produce Recurring Reports on Schedule for how the checks fit the production process. To have your reporting data assessed, request a data-quality audit.

Sources

• Gartner, Magic Quadrant for Data Quality Solutions (July 2020), survey of 154 enterprise customers: average $12.9 million annual cost of poor data quality.
• IBM Institute for Business Value (2025): 43% of chief operating officers rank data quality as their top data priority; over a quarter of organizations estimate losses above $5 million a year.
• Harvard Business Review (Nagle, Redman and Sammon, 2017): 47% of newly created data records contain at least one critical error; only 3% of companies’ data met basic quality standards.
• Harvard Business Review (Redman, 2016): knowledge workers waste about 50% of their time finding and correcting data.
• Intuit Enterprise Suite, Future of Finance 2026 Report (2,000 finance leaders, May 2026): 70% report no single source of truth.

Numbers are cited from their sources and dated. Where a source is a vendor benchmark, the sample size is stated. Verify figures against the primary source before republication.