When a mix shift explains a metric drop, do you report the raw or the mix-adjusted number to leadership?

## Two numbers answering two different questions When a blended metric moves and the decomposition shows the movement is mostly composition, you end up holding two figures: - The **raw (crude)** metric: what the blend actually did. It reflects the business as experienced — the money collected, the servers needed, the support tickets received. - The **mix-adjusted (standardised)** metric: what the blend would have done had the segment composition stayed at a fixed reference. It isolates within-segment performance. Neither is more honest than the other in the abstract. They answer different questions, and the failure mode is presenting one while the audience believes it is hearing the other. ## Why the raw number must lead The adjusted number is counterfactual: it describes a composition that did not occur. Nobody can spend it, staff against it, or forecast from it. A leadership review that opens with "conversion is up 1.0 points on a constant mix" while the company actually converted 1.4 points worse has told the audience something true and left them with a false belief. That is the specific failure worth being blunt about — adjusted-first reporting is how organisations get surprised by their own P&L. So: raw as the headline, adjusted immediately underneath as the explanation, and a reconciliation that ties the two together. "Blended conversion -1.4 points; within-segment +1.0, mix -2.4" is one line and closes the gap. ## The ownership question is the real judgment call Mix effects fall into two very different categories, and treating them alike is where leaders go wrong. **Mix as a decision.** Marketing raised spend on a cheaper channel; the company launched in a lower-ARPU country; a new low-priced tier drew volume. Here the mix shift **is** the outcome of a deliberate bet. Normalising it away deletes the very thing that should be evaluated — and evaluated on its own terms, since cheaper traffic converting worse can still be the better trade on cost per acquisition. The mix term belongs on the deciding team's line, priced in the currency that decision was made in. **Mix as exogenous drift.** Seasonal composition, a platform's own audience shifting, a partner changing its traffic. Nobody chose it, and holding the funnel team to a blended number that this drift dominates is a misattribution. Here the adjusted view is the fairer scorecard. The organisational discipline is to say which of the two you believe it is, and to name the owner. "The mix moved and therefore nobody is accountable" is the answer to avoid — mix shifts have causes, and usually a team. ## Governance of the adjustment An adjustment is only credible if it cannot be tuned. Three rules do most of the work: 1. **Fix the reference mix in advance** and keep it constant across periods. A standard recomputed each quarter makes the time series incomparable and hands the analyst a dial. 2. **Fix the segmentation too.** Which dimension you adjust on (channel, country, device, tenure) changes the size of the mix term. Choosing it after seeing results is the same defect one level up. 3. **Require exact reconciliation.** The adjusted view and the mix term must sum to the raw delta. A decomposition that does not tie out is a story, not an analysis. Without these, "it was mix" becomes the universal explanation for every miss, and the credibility of the whole analytics function erodes — the second time a mix explanation precedes a bad quarter, nobody believes the third. ## Incentives: what happens when the adjusted metric becomes a goal Put a mix-adjusted rate in a team's objectives and you have told them that composition is not their problem. Predictably, they optimise within-segment rates and become indifferent to a mix they influence — a funnel team stops caring which traffic arrives, an acquisition team stops caring what it converts to. The company then grows a portfolio nobody chose. The usual remedy is to pair the metrics rather than pick one: the adjusted rate as the performance measure, plus a **mix guardrail** (share of volume from the low-value segment, or the absolute total the mix feeds) that fails loudly if composition drifts past a threshold. And keep at least one absolute total in the goal set, because absolutes cannot be improved by re-weighting anything. ## What a strong answer sounds like Refuse the either/or framing; specify the order and the reconciliation; separate chosen mix from exogenous mix and assign an owner for each; pre-commit the standard and the segmentation; and close on the incentive consequence of whichever metric ends up in the goal document. The arithmetic is the easy half — the question is really about what a number does to an organisation once it becomes a target.