The Difference Between Correlation and Causation in Growth Reporting

When revenue increases during a period of active marketing, most analysts conclude that the marketing worked. When it decreases, most conclude that something in the marketing failed. Both conclusions rely on the same assumption: that the data available is sufficient to draw causal inferences. It is not.

Standard growth reporting is almost entirely correlational. It shows what happened at the same time as revenue moved. Channel X was active during a period of growth. Spend increased in quarter three, and so did revenue. The report records the co-occurrence and implies the cause. The implication is rarely examined.

The structural limitation is this: standard analytics captures what happened inside the funnel. It does not capture what was happening in the market at the same time. A business might increase paid search spend and observe a revenue increase. But if category-level search demand also increased during that period — driven by a seasonal trend, a regulatory change, or a competitive event — the revenue increase was partially or wholly caused by external demand, not by the paid activity.

Attribution models attempt to resolve this, but they operate on the same closed population: visitors who arrived. They assign credit for conversion across touchpoints within the funnel. They cannot assign credit — or absence of credit — to the demand that never entered the funnel at all.

The practical consequence is that growth decisions are made on incomplete evidence. If marketing spend is credited for a revenue increase that was primarily caused by a market tailwind, the organisation will increase that spend expecting to replicate the result. When the tailwind subsides, the result will not replicate, and the spend will appear to have failed.

Causal measurement requires a counterfactual. What would have happened without the intervention? Standard analytics cannot answer this. But demand-level data from Google Search Console can establish a baseline: what was the total category demand in the market, what was the capture rate before the intervention, and what changed after it?

That is not a complete causal analysis. But it is the minimum necessary condition for one. Without knowing what demand existed and what proportion was captured, any attribution of revenue to marketing activity is a correlation presented as a cause.

The decisions that follow from that error are not minor. They determine where capital is allocated, which channels are expanded, and which are cut. Getting the causal inference wrong at this level is not an analytical nicety — it is a structural source of capital misallocation.