Make the Company Readable

Senior leaders must work through representations. The risk begins when an approved representation cannot be tested against the system it describes.

No director can understand every service or supplier, nor every process and data flow, in a large company. Abstraction is part of governing. The practical requirement is that a material claim can lead to the mechanism and evidence it summarises, allowing the representation to remain concise without becoming independent of reality.

The finance director can connect margin to cost and revenue in the monthly accounts. She cannot connect either to the system behaviour her money changed. Technology reports delivery and incidents, Operations reports volumes and edge cases, and Risk reports controls. Each account may be accurate while their meanings of success describe different companies.

Composite

On a Tuesday, non-executive director Helen Ward asks what “91 per cent straight-through processing” means in the board pack. Eleven days later she has three answers: 91 per cent of applications that reached the decisioning service, 62 per cent of applications submitted, and 44 per cent of customers who started. Each number was produced correctly. Nobody had reconciled them because, until she asked, no decision had depended on which one was true.

The represented company is the company described by board packs and functional reports. The running company is expressed by actual decisions, software, operational work, suppliers, data flows, controls, and customer outcomes. Leadership needs the first because the second is too large to inspect directly; the risk appears when discrepancies can be removed from the representation without being removed from work.

You should ask questions that cross Finance, Technology, Operations, and Risk reports: which process changed for the money spent, which result moved, and which decision caused the delay? When no owner can answer, a translator usually makes the accounts agree well enough for the meeting.

A cross-functional account often survives as a portfolio workbook mapping Finance's project codes to Technology's service names and Operations' process names. Programme managers and portfolio analysts maintain that synthesis with delivery leads, often supported by business partners. It remains fragile because the person reconciling the meanings has no right to change the boundaries generating them, and the map may leave with its maintainer.

Representations depend on people reconciling them.

Finance, technology, and operations produce separate representations. Translation roles absorb contradictions before the combined narrative reaches the running system.

Helen Ward's question does not need the support of a famous failed programme. A public failure may show that a readiness representation became detached from evidence, but it cannot by itself establish why the organisation was designed that way or which correction would have worked. Her three correct numbers expose the governing problem inside an ordinary company: a claim can be locally valid and still fail to describe the complete outcome.

Making every material claim traceable is modest to state and difficult to fake. If you write the claim, you should be able to walk a challenger from the summary to the process and system state, then the exact decision and record behind it. A board still receives an abstraction, while the person preparing it can no longer close a gap by choosing the most reassuring of several incompatible meanings.

You need not place source code in a board pack. A claim that automated renewals reduced handling time can bind itself to the process definition and deployed version for a declared population and watch window. It also names the edge cases and the owner of the conclusion. Directors receive the summary; the people preparing and challenging it can reproduce the comparison. When they cannot, the uncertainty belongs in the claim instead of in a private conversation after the meeting.

Six weeks later, straight-through processing returns to Helen Ward's board pack at 62 per cent of submitted applications. A footnote names Operations as the definition owner and lists the excluded edge cases. Helen asks the same question. This time the report's author can trace the percentage from the sentence to the decisioning service's event stream in one call. The underlying systems remain unchanged, but the claim now lets a director tell which company the number describes.

A testable representation remains an abstraction, but it can no longer hide which company it describes.

Chapter 16 turns claim traceability into a reader with a bounded assignment that tests governed claims against their cited implementation and production evidence.