Finance data quality and a single source of truth

# Finance data qualitydata qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.Voir la définition complète → and a single source of truth

In 2015, Hertz restated three years of financials and eventually disclosed $235 million in accounting errors. The post-mortem was not a story of fraud masterminds or exotic derivatives. It was a story of reconciliations that didn't reconcile, subledgers that didn't agree with the general ledger, and a control environment where "the number" depended on who you asked. The CFO's most expensive discovery was mundane: the company had no single, governed version of its own dataown dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.Voir la définition complète →.

Every CFO has lived a smaller version of this. Sales reports $4.2M in bookings; the GL shows $3.9M in revenue; the board deck rounds it to "roughly $4M." Three "true" numbers, three owners, zero reconciliation. The instinct is to blame the report. The report is fine. The problem is upstream, in the data model that feeds it. This lesson is about diagnosing reporting pain as a data-architecture problem and engineering the fix: a governed chart of accounts and a single source of truth that forces every report to tie out.

Why reporting pain is almost always a data problem

When a report is wrong, finance teams instinctively fix the report. They patch the formula, adjust the pivot, add a reconciling line called "other." This treats the symptom. The disease is that the same business event is being represented differently in different systems, and no authoritative layer arbitrates between them.

Consider the anatomy of a single transaction, say, an enterprise SaaS deal. It exists in the CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.Voir la définition complète → as a booking, in the billing system as an invoice schedule, in the GL as deferred and recognized revenue, in the data warehousedata warehouseA central repository that consolidates data from many source systems into a structured, query-optimized store designed for analytics, reporting, and business intelligence.Voir la définition complète → as an ARRARRAnnual Recurring Revenue (ARR) is the normalized, predictable revenue a subscription business expects to earn from active contracts over a single year.Voir la définition complète → record, and in the board deck as a growth metric. Five systems, five representations, five potential definitions of "the customer," "the amount," and "the date." If any two disagree on grain, timing, or hierarchy, your reports will never tie out, and no amount of Excel heroics will permanently fix it.

The strategic reframe for the CFO is this: you are not running a reporting function, you are running a data supply chain. Raw events enter at the source, get transformed through defined rules, and exit as finished goods (reports). Like any supply chain, quality is determined at the point of manufacture, not at inspection. A number that arrives wrong at the warehouse cannot be made right in the board deck.

This yields a practical diagnostic. When a report is questioned, ask three questions in order:

1. Definition, Do all parties agree on what the metric *means*? (Is "revenue" gross or net of refunds? Is a "customer" a legal entity or a billing account?)

2. Lineage, Can you trace the number from the report back to the originating transactions, hop by hop?

3. Reconciliation, At each hop, does the total in equal the total out, with variances explained?

If you cannot answer all three within an hour, you do not have a reporting problem. You have a data-governance vacuum. The most valuable thing a CFO can install is not a better BIBITechnologies and processes that turn raw data into actionable insights via reporting, dashboards and analysis, so teams can decide based on facts rather than intuition.Voir la définition complète → tool, it's the discipline that makes these three questions answerable on demand.

The cost of a fragmented model

Fragmentation is quietly expensive, and the costs are rarely on any P&L line. There's the close tax: teams burning days reconciling systems that should agree by design. There's the credibility tax: an analyst or director catches a discrepancy, and now every number you present is discounted. And there's the decision tax, the worst of the three, where capital gets allocated on numbers that are subtly wrong, and no one knows until the variance shows up in cash.

The CFO's job is to make the case that consolidating on a single source of truth is not an IT convenience project. It is a risk-reduction and decision-quality investment with a measurable return: fewer restatements, faster closes, and executives who trust the dashboard enough to act on it without re-deriving it.

Building the governed chart of accounts

The chart of accounts (COA) is the DNA of your financial data model. Most CFOs inherit a COA that grew like a coral reef, accretive, unplanned, and structurally rigid. Someone needed to track a new expense in 2019, so they added account 6847. Now you have 2,400 accounts, forty of which are used, and a "miscellaneous" bucket absorbing anything ambiguous. This is where data qualitydata qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.Voir la définition complète → goes to die.

A modern COA is deliberately engineered around a critical principle: separate the natural account from the dimensions. The natural account should answer only *what kind of thing* this is, salaries, software, revenue, receivables. Everything else, which department, which entity, which product, which region, which project, belongs in dimensions (also called segmentssegmentsDividing a market into distinct groups of customers who share similar needs, characteristics or behaviours, so each group can be served with a tailored approach.Voir la définition complète → or tags), captured alongside the transaction, not baked into the account number.

The failure mode is the "smart" account number, where account 6100-US-SALES-EMEA encodes the type, entity, function, and region into one string. It looks efficient. It is a trap. When you reorganize, acquire, or enter a new market, you must either create dozens of new accounts or misclassify transactions into the old ones. A dimensional model absorbs organizational change gracefully; a smart-coded COA fossilizes your 2021 org structure into your ledger forever.

Design principles for a reengineered COA

When you rebuild, and most CFOs should, at least once, ideally alongside an ERP migration, apply these constraints:

• Every account has one unambiguous definition. Write it down. If two people can book the same transaction to two different accounts, you have a definition gap that will surface as unexplained variance.
• Minimize the natural-account list; maximize dimensional richness. A clean COA might have 200-400 natural accounts. Analytical depth comes from dimensions, which can be combined infinitely without new accounts.
• Design dimensions to survive reorganization. Cost center, legal entity, product line, geography, and customer segment should be modeled as independent axes so you can re-slice history when the org changes.
• Enforce mutually exclusive, collectively exhaustive (MECE) structure. No transaction should have two valid homes; every transaction should have exactly one.
• Kill the "miscellaneous" account. A material "other" line is a confession that your model has a gap. Force classification at entry.

The payoff is what practitioners call write-once, report-many. You capture a transaction once with full dimensional context, and every downstream report, by entity, by product, by region, by function, derives from that single record. The management view and the statutory view become two projections of the same data, not two separately maintained datasets that mysteriously never agree.

Engineering the single source of truth

A governed COA gives you clean *inputs*. A single source of truth (SSOT) is the architecture that ensures every *output* derives from one authoritative dataset. The two must be built together; a beautiful COA feeding six ungoverned spreadsheets accomplishes nothing.

The SSOT is not a single database or a single tool. It is a governed layer, typically a data warehousedata warehouseA central repository that consolidates data from many source systems into a structured, query-optimized store designed for analytics, reporting, and business intelligence.Voir la définition complète → with a defined semantic model, that sits between your source systems and your reports. Its defining property: every metric that appears in any report is calculated once, in one place, according to one definition, and every consumer pulls from that layer. The CFO's dashboard, the board deck, the FP&A model, and the sales leaderboard all resolve "net revenue" to the same query.

The semantic layer and metric governance

The technical instrument that makes this real is the semantic layer, a governed dictionary of metric definitions that lives above the raw data. Instead of "net revenue" being redefined in each spreadsheet, it is defined once, in code, as: *gross revenue minus refunds and credits, recognized per ASC 606, by transaction date.* Every tool references that definition. Change it once, and it changes everywhere, with a version history.

This shifts the CFO's governance role. You stop policing individual reports and start owning a metrics catalog: the authoritative list of every KPIKPIKey Performance Indicator, a measurable value that shows how effectively you're achieving a specific objective, tracked over time against a target.Voir la définition complète → the company reports, its precise definition, its owner, and its data lineagedata lineageData lineage maps how data moves and transforms across systems, from origin to consumption, showing where it came from, what changed it, and where it goes.Voir la définition complète →. When someone proposes a new metric, it enters through a governance process, defined, reconciled, and approved, rather than being invented in a deck the night before a board meeting.

Reconciliation by design, not by heroics

The deepest quality mechanism is architectural: build the system so that reports *cannot* diverge from the ledger without triggering an alert. This is reconciliation by design.

In practice, install automated controls that continuously check for agreement across the data supply chain:

• Subledger-to-GL tie-outs that flag any variance in real time, not at month-end.
• Source-to-warehouse row counts and control totals so you know when a data pipelinedata pipelineETL (Extract, Transform, Load) is a data integration process that pulls data from sources, reshapes it into a consistent format, and writes it into a target system.Voir la définition complète → silently drops records.
• Cross-system reconciliation (CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.Voir la définition complète → bookings to billing to GL revenue) with defined, monitored bridges between each definitional gap.
• Data-quality tests on the pipeline itself, no null customer IDs, no negative quantities, no dates in the future, no orphaned records.

The cultural shift this enables is significant. When reconciliation is automated and continuous, your close accelerates because you are not discovering breaks in the final days, you fixed them the day they occurred. Hertz's failure was, at root, the absence of this: reconciliations that were manual, periodic, and therefore skippable under deadline pressure.

Assigning ownership: data governancedata governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.Voir la définition complète → in practice

Architecture without accountability decays. The SSOT requires named owners:

• Data owners, business leaders accountable for the definition and quality of a domain (the VPVPA clear statement of the benefits your product delivers, the problems it solves and why customers should choose you over alternatives.Voir la définition complète → of Sales owns "booking"; the Controller owns "recognized revenue").
• Data stewards, operators who enforce the rules day to day and triage quality alerts.
• A data governance council, ideally chaired or sponsored by the CFO, that arbitrates definition disputes and approves changes to the metrics catalog.

The CFO's role here is not to write SQLSQLSales Qualified Lead: a prospect the sales team has validated as ready for direct outreach and a proposal, having passed clear qualification criteria.Voir la définition complète →. It is to insist that no metric is orphaned, that every definition has a single accountable owner, and that changes flow through governance rather than through side channels. This is where finance controllership and modern data management converge, and increasingly, where the CFO's mandate genuinely lives.

Making it real: the CFO's sequencing problem

The trap is treating this as a two-year platform rebuild that delivers value only at the end. It won't survive the budget cycle, and it shouldn't. The discipline is to sequence for early, visible wins while building toward the target architecture.

Start with your top ten reported metrics, the ones the board, lenders, and executive team actually use. Define each precisely, trace its lineage, and reconcile it end to end. This alone surfaces most of your definitional debt and delivers immediate credibility. You will almost certainly find that two "trusted" numbers were never reconciled to each other.

Next, govern the COA before the next ERP migration or major acquisition, not after. These events are the natural, and often only, windows to reengineer the account structure without a standalone change-management war. A CFO who walks into an ERP implementation without a redesigned, dimensional COA has surrendered the single best chance to fix the foundation.

Then build the semantic layer incrementally, one metric domain at a time, revenue first, then cost, then cash, retiring the shadow spreadsheets as each domain becomes authoritative. Track a hard metric: the percentage of board-reported numbers that pull directly from the governed layer. Drive it toward 100%.

Finally, resist the temptation to boil the ocean. A pristine model of the ten metrics that drive decisions beats a comprehensive model of two hundred that no one trusts. Data qualityData qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.Voir la définition complète → is not a state you reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.Voir la définition complète →; it is a discipline you maintain, and the CFO's lasting contribution is installing that discipline into the operating rhythm of the finance function.

Key Takeaways

• Diagnose reporting failures as data-architecture failures. When a number is questioned, run the definition, lineage, reconciliation test before you touch the report. If you can't answer all three in an hour, you have a governance gap, not a formula bug.
• Separate natural accounts from dimensions. Keep the account list short and unambiguous; push all organizational context, entity, department, product, region, into dimensions so your model survives reorgs, M&A, and market entry without rework.
• Build one governed semantic layer where every metric is defined once. Own the metrics catalog personally: every KPIKPIKey Performance Indicator, a measurable value that shows how effectively you're achieving a specific objective, tracked over time against a target.Voir la définition complète → needs one definition, one owner, and traceable lineage. No metric enters the board deck without passing through governance.
• Engineer reconciliation by design. Automate continuous subledger-to-GL and cross-system tie-outs so breaks surface the day they happen. This is what accelerates the close and prevents the Hertz scenario, not more year-end heroics.
• Sequence for early wins. Start with the top ten decision-driving metrics, fix the COA during the next ERP or acquisition window, and measure progress by the share of reported numbers that pull directly from the governed source.