Skip to content

Data

Management reporting in a group of companies: one set of numbers

How a group of companies gets one set of numbers for management reporting: a metric dictionary, shared reference lists, nightly checks and a weekly business review.

Published
9 min read
9 min read

In a group of companies each business counts revenue, active customers and margin in its own way, so every board meeting starts with reconciling numbers. Another reporting tool does not cure this. Four agreements do: one definition for each metric with a named owner, shared reference lists, automatic checks, and one weekly review that everyone attends.

In short

  • Gartner puts the average cost of poor data quality at 12.9 million US dollars a year for an organisation, on research from 2020.
  • The barrier is people and process. In the Wavestone survey of 2024, 77.6% of data leaders named culture, people and process as the main obstacle.
  • Statutory consolidation under IFRS 10 already demands uniform policies, the removal of intragroup flows and one reporting date. Management reporting needs the same discipline every week.
  • Start with a short list of metrics. Each one gets a definition, a formula, an owner and a source.

Why each company’s numbers differ

Companies in a group usually run different systems: one accounting package here, another there, a booking system, a CRM, spreadsheets. Each team then builds its own reports on top. Gartner names inconsistency across sources as the hardest data quality problem and gives the cause: data "stored and maintained in silos with significant overlaps, gaps or inconsistencies". The same page gives the price. Poor data quality costs organisations at least 12.9 million US dollars a year on average, a figure Gartner dates to its research from 2020.

Manual work makes it worse. Gartner surveyed 497 people who work in controllership and published the results on 21 February 2024: 18% of accountants make financial errors at least daily, and 59% make several a month. Gartner’s analyst warned that when such errors reach the monthly close, the company "may make business decisions based on incorrect data".

The close itself is slow. The APQC benchmark across 11,223 companies puts the median time from the first trial balance to agreed monthly consolidated statements at six calendar days. A report that arrives a week after the month ends cannot steer the month.

Some estimates go higher. Thomas Redman wrote in MIT Sloan Management Review in 2017 that bad data costs most companies 15% to 25% of revenue. That is an expert’s estimate and not a measurement, but it shows the order of magnitude.

The problem is rarely the software. In the Wavestone survey of data and AI leaders, 2024 edition, only 37% said their efforts to improve data quality had succeeded, and 77.6% named culture, people, process and organisation as the main barrier to becoming data-driven.

What statutory consolidation already teaches

A group that reports under IFRS already knows how to produce one set of numbers. IFRS 10 presents a parent and its subsidiaries as a single economic entity and spells out how to get there:

IFRS 10What it requiresThe same rule in management reporting
Paragraph 19Uniform accounting policies for like transactionsOne definition of each metric for all companies
Paragraph B87Adjustments when a group member uses other policiesA documented conversion when a company counts differently
Paragraph B86(c)Intragroup income, expenses and cash flows are eliminated in fullSales between group companies are removed from group revenue
Paragraph B92The same reporting date for the parent and its subsidiariesOne calendar: the week and the month end on the same day everywhere

The difference is frequency. Statutory consolidation is done once a quarter or once a year, and it takes weeks. Managers need the same quality every Monday.

There is a deadline as well. IFRS 18, issued in April 2024, applies to annual periods that begin on or after 1 January 2027. It requires companies to disclose management-defined performance measures, the subtotals of income and expenses that management uses in public communications. For a group that reports under IFRS, the definitions of its own metrics stop being an internal matter.

The metric dictionary

The first agreement is a metric dictionary: a list of the key metrics in which each one has a definition, a formula, a level of detail, an owner and a source. An illustrative example:

MetricDefinitionLevel of detailOwnerSource
RevenuePaid orders net of returns and VAT, without sales to other group companiesCompany, dayGroup CFOAccounting system
Active customerA customer with at least one paid order in the last 90 daysCompany, monthHead of salesCRM and payments
Gross marginRevenue minus the direct cost of goods and delivery, divided by revenueCompany, product category, weekCFO of each companyAccounting system

The definitions are for each group to choose. What matters is that there is one of each. The Kimball Group’s rule for conformed facts says it precisely: if the same measure appears in different places, its definition must be identical before the figures are compared or added, and if the definitions are incompatible, the measures "should be differently named to alert the business users and BI applications". Two companies that mean different things by revenue need either one agreed definition or two names.

Where the definition lives matters too. When each dashboard carries its own formula, the formulas drift apart. The alternative is to define each metric once, in code, in a layer that every tool reads. The documentation of the dbt Semantic Layer describes the effect: "If a metric definition changes in dbt, it’s refreshed everywhere it’s invoked." dbt is one example of this approach, not the only one. In its model, metrics are built from simple ones into ratios and derived metrics, so margin is computed from the same revenue everywhere.

An owner for each metric is what the textbooks call data governance. The DAMA body of knowledge defines it as "the exercise of authority and control (planning, monitoring, and enforcement) over the management of data assets", in the wording reproduced by the DAMA Rocky Mountain Chapter. In practice it means a named person who decides what the metric means and who answers when it looks wrong.

Shared reference lists and a plan on one page

The second agreement is about the lists that describe the business: the calendar, the company, the customer, the product category, the location. Kimball calls them conformed dimensions. Dimensions conform when "attributes in separate dimension tables have the same column names and domain contents". Only then can results from different processes be lined up in one report, which the same page calls "the essence of integration". The lists are defined once, together with the people in the business who answer for the data, and then reused.

For a group this is where synergies become visible. Only with one definition of a customer can anyone see that the clients of one company also buy from another.

The same school offers a planning tool that fits on one page, the bus matrix: "The rows of the matrix are business processes and the columns are dimensions." Sales, payments, bookings and deliveries go in the rows, and the shared lists go in the columns. The matrix is then built one row at a time, not as one large project.

Checks before anyone opens a dashboard

The third agreement is that machines check the numbers every night, before people see them. Data quality checks look for gaps, duplicates, values outside the allowed list, stale data and unusual volumes. If a load fails, the team learns it in the morning and not in the meeting.

Gartner notes on its data quality page that business leaders agree that data quality matters but "do not view it as their responsibility". Checks without owners change nothing. Each failed check needs a person who is expected to react.

The weekly business review

The fourth agreement is a meeting. The best-described example is Amazon’s weekly business review, known from the book Working Backwards by former Amazon executives Colin Bryar and Bill Carr. Cedric Chin of Commoncog, who learned the practice from Bryar, describes the rhythm. The metrics for the past week are generated after midnight on Sunday, each metric owner reviews them on Monday, departments hold their reviews on Tuesday, and the company review takes place on Wednesday. The meeting lasts 60 minutes. Finance tracks the goals and can audit any metric independently.

For a group, the useful part is the list of what makes the weekly business review work:

  • the same pages every week, in the same order;
  • an owner for each metric who explains the change;
  • finance as the neutral keeper of the numbers;
  • a standard chart that shows the last six weeks next to the last twelve months.

The book also records how the review fails. In one case, quoted by Commoncog, "the number of metrics we were trying to track kept ballooning", and the lesson was that the senior leader should have "limited the metrics to be reviewed to a specific, essential set". Amazon’s own leadership principles give the purpose in one line: leaders "stay connected to the details, audit frequently, and are skeptical when metrics and anecdote differ".

A realistic first eight weeks

None of this needs a finished platform. The sequence we use looks like this:

  1. Weeks 1 and 2: understand. Interviews with leaders, a list of the decisions that matter, an audit of systems and data, and a baseline for the result.
  2. Weeks 3 and 4: agree. Metric definitions, a staged plan and an architecture with costs.
  3. Weeks 4 to 8: first value. A first management dashboard and a first weekly review on a thin slice of data.
  4. From month 2: build and run. The full foundation, company by company.

Where it goes wrong

  • Too many metrics. A review of everything is a review of nothing. Start with the short list that leadership actually decides on.
  • No owner. A metric without an owner drifts until two teams report two versions.
  • Definitions inside dashboards. Each new dashboard then becomes a new version of the truth.
  • AI before order. Gartner surveyed 1,203 data management leaders and reported on 26 February 2025 that 63% of organisations either lack the right data management practices for AI or are not sure they have them. It predicts that through 2026 organisations will abandon 60% of AI projects that are not supported by AI-ready data. We look at that side in the article on AI agents in a holding company.

How Sahman approaches this

Sahman is at an early stage. We start every client with the metric dictionary, define each metric once in code, check the data every night and run the weekly business review with an embedded analyst. We will publish results only after we have measured them. The architecture is described on the platform page.

Sources

  1. 1Data Quality: Best Practices for Accurate Insights. Gartner
  2. 2Gartner Survey Shows That a Third of Accountants Make Several Financial Errors Per Week due to Capacity Constraints. Gartner, 2024-02-21
  3. 3Cycle time in days to complete the monthly consolidated financial statements. APQC, Open Standards Benchmarking
  4. 4Seizing Opportunity in Data Quality. Thomas C. Redman, MIT Sloan Management Review, 2017-11-27
  5. 52024 Data and AI Leadership Executive Survey. Executive Summary of Findings. Wavestone, 2024
  6. 6IFRS 10 Consolidated Financial Statements, as endorsed by Commission Regulation (EU) No 1254/2012. EUR-Lex, 2012-12-11
  7. 7IFRS 18 Presentation and Disclosure in Financial Statements. IFRS Foundation, 2024
  8. 8Conformed Facts. Kimball Group
  9. 9dbt Semantic Layer. dbt Labs, 2026-09-29
  10. 10Creating metrics. dbt Labs, 2026-07-20
  11. 11DMBoK Figure 14 Context Diagram: Data Governance and Stewardship. DAMA Rocky Mountain Chapter, 2023-08-23
  12. 12Conformed Dimensions. Kimball Group
  13. 13Enterprise Data Warehouse Bus Matrix. Kimball Group
  14. 14Working Backwards: Insights, Stories, and Secrets from Inside Amazon. Colin Bryar and Bill Carr, St. Martin’s Press, 2021-02-09
  15. 15The Amazon Weekly Business Review (WBR). Cedric Chin, Commoncog, 2024-07-02
  16. 16Leadership Principles. Amazon
  17. 17Lack of AI-Ready Data Puts AI Projects at Risk. Gartner, 2025-02-26

This article is also available in Azərbaycanca, Русский

Want to see how this works on your group’s data?

A pilot takes four weeks, covers one question your leader cares about and has a written measure of success.