The Number Was Right, and Nobody Could Act On It

October 19, 2026

Business Intelligence

Consider a retailer whose analysts find that a regional format loses money once shared costs are allocated. The data may be sound, but the decision about closing or changing the format sits with a different executive.

The team can trace the lineage, agree the definitions and check the assumptions. That builds confidence in the finding, though it says nothing about who decides what happens next. If the finding challenges the regional director’s own earlier decision, it needs a route to an independent review. Analysts can publish the result, and finance can validate it, but a decision maker still has to weigh it against customer, legal and operational consequences.

The data was governed. What nobody had governed was the route from that finding to a business decision.

A useful distinction from data governance

Vijay Khatri and Carol V. Brown separated governance from management in Communications of the ACM in 2010, building on earlier IT governance work. Their paper is about decision rights for data: which data decisions have to be made and who makes them. Governance assigns accountability for those decisions, and management makes and carries them out.

The same logic extends one step downstream. A decision to change a store format has its own authority and evidence requirements, and linking the two is a matter of organisational design.

A data programme can rightly concentrate on stewardship, quality, definitions, lineage and access. If it also promises better decisions, it needs to say how a material finding reaches the person who owns the business choice, including when the finding concerns that person’s own earlier decision. The data work stays necessary. The extra test is whether its output reaches a forum that can weigh costs, trade-offs and a decision date, because an accurate number can otherwise circulate without anyone being accountable for a response.

The convenient number and the inconvenient one

Decision rights show up most clearly when a finding challenges a current plan. A favourable number lets the team carry on. A difficult one may call for a pricing change, a service redesign or the closure of a format. Each of those has different owners and consequences, so the escalation route should be mapped to the decision at stake.

A quality check tests the number. A decision review tests what the organisation does with it. A programme that assigns production and validation roles can still leave the business decision with the existing hierarchy, which often works well. The risk begins when the person whose decision is being challenged also controls whether the evidence reaches a wider review. If that person disputes the evidence, the team needs a stated way to resolve the dispute and record the outcome.

Why the org chart does not answer this

An organisation chart usually shows who can change a regional format. It rarely shows how conflicting evidence reaches that person, what independent review is available or when a decision is due.

A director can raise legitimate methodological questions about an unwelcome finding. Agreeing the evidence standard before any specific dispute makes those questions easier to answer and stops them turning into indefinite delay. The decision route should cover challenge and response, with a second reviewer wherever a material conflict of interest could affect the assessment.

What has to be assigned before the number arrives

Agree the route before the next high-impact finding, because a live dispute is a poor moment to negotiate authority and standards of proof from scratch. For the few business decisions most likely to follow from the data, put three things in writing.

The first is the class of decision the finding could inform, such as pricing, a store format or a supplier review. The second is the role that must respond, the deadline for that response and a route to independent review when the finding challenges the owner. Existing governance bodies can fill that role if their remit includes it. The third is the evidence needed for a decision, including checks on cost allocation, sensitivity and wider commercial effects, with the threshold recorded and any later change to it explained.

Test the route against one earlier finding that caused debate. Reconstruct who saw it first, what questions came up, when a decision owner received it and which evidence settled the matter. The exercise may show a sound process that only needs a deadline, or a material handoff with no owner. Either result is more useful than drafting an elaborate charter before anyone has mapped the decision.

The decision record should keep more than the approval. It should state what the analysis found, the assumptions that would change the finding, the commercial options considered, and the reason for the choice. An executive may reasonably keep a loss-making format open because it supports another part of the business. A record of that judgement lets the next review tell a deliberate trade-off from an unresolved handoff, and it shows the data informed the decision even where its recommendation was declined.

Take a disputed cost allocation. A second reviewer can test alternative methods and show how each changes the conclusion. The original result may hold across all of them, or the decision may turn out to rest on an uncertain assumption. Either outcome helps the executive responsible. The standard of evidence should allow a clear decision under uncertainty, with a date for revisiting it. Waiting for perfect certainty can leave a costly format open indefinitely, and a rushed response can close something the business still needs.

A data team should also be clear about what it does not own. Analysts can explain the calculation and its sensitivity, finance can check the accounting, and an operating leader can assess customers, staff and practical options. Senior management decides which trade-off the business will accept. Naming those contributions protects analysts from being held responsible for a closure they cannot authorise, and it protects executives from receiving a number stripped of its context.

At the next planning meeting, the same decision record should show whether the earlier choice produced the expected result. That gives the business a way to learn from an unpopular decision or a well-intentioned delay, and it gives the data programme a measure of use beyond the number of reports it issued.

The bureaucracy objection

A new decision route adds process, so its scope and response times should match the value and risk of the decision it governs. The extra step earns its cost when it stops a finding circulating without an owner. It slows a simple decision, so reserve it for cases where the cost or the conflict warrants it.

Start with a few material decisions, such as pricing and portfolio choices, and test the route on real findings, since the right scope depends on the business. Record how long decisions take and whether the analysis reaches the authorised owner. Routine findings can follow established management processes, with a check that someone is accountable for acting wherever the data has a material consequence.

The part that is uncomfortable on purpose

Writing the route is a small task next to agreeing it. Finance, operations and the affected business owners have to settle who decides, what can be challenged and how a disagreement is resolved.

The agreement costs the regional director sole control over whether a finding about their own format reaches a review, and it costs senior leaders a written response by a set date. Those are the concessions behind a data programme’s promise that better information will change decisions.

In the retailer scenario, the requirement is specific: the loss finding reaches an authorised review before another planning cycle passes.