When three medicines numbers can all be right

Conflicting figures are often a definition problem before they are a data problem.

You can put three credible medicines figures on one page and create the impression that only one can be true. In practice, each may describe a different population, price basis, time period or stage of the pathway. The danger begins when a team selects the most convenient number without preserving those definitions. A decision that looks precise can then rest on a comparison that was never valid. The better future is not a single perfect figure; it is a decision record that shows exactly what each figure means.

Start with the denominator, not the headline.

Before comparing values, ask who or what is counted. Is the denominator all people with a condition, those clinically eligible under a defined criterion, those currently treated, or those whose records contain enough information for review? The same discipline applies to activity and spend. National totals, local expenditure, acquisition cost, modelled resource effect and realised cash movement are different measures. Labelling them consistently removes much of the apparent disagreement before anyone debates the conclusion.

Give every number a four part identity.

A useful medicines number should travel with four things: its source, definition, effective date and decision purpose. Add the transformation history when a calculation has been applied. That small data contract lets another person reproduce the figure and decide whether it is fit for the question in front of them. It also makes updates safer. When a tariff, population estimate or policy position changes, the team can see which conclusions depend on the old input instead of rebuilding the analysis from memory.

Reconciliation should preserve difference rather than average it away.

The aim is not to force every source into one blended answer. It is to explain why the values differ and which one belongs in each decision. A national planning estimate may be appropriate for strategic capacity. A locally verified cohort may be necessary for operational work. A modelled opportunity can justify further analysis without becoming a realised saving. A good reconciliation table keeps these lanes separate and records uncertainty, exclusions and sensitivity rather than allowing a single attractive total to dominate the conversation.

The accountable question is what the number authorises.

A number can trigger investigation, support prioritisation or inform an authorised decision; those are not the same status. Teams should state the next action the figure is permitted to support and the evidence required before moving further. This is particularly important when clinical, financial and operational lenses point in different directions. Visible authority and review thresholds protect both urgency and caution: promising signals can move, while untested assumptions cannot quietly become commitments.

The evidence boundary applies to both publications.

The original commentary and this later reflection are founder views. They do not demonstrate that PHARMORIS has reconciled live institutional data, produced savings, been deployed or been adopted. What they establish is the method I expect the company to follow: define first, reconcile second and only then decide. The archived source record retains the publisher’s original byline.