Beware the Rosser Reeves effect when interpreting tracking data

People who remember the advertising buy more. That finding looks causal because it points in the direction everybody has been paid to hope for. The reverse story is less celebratory and equally plausible: buyers notice the advertising because they already know, use or prefer the brand.

Existing customers recognise familiar packaging and messages. Frequent shoppers notice retailer communication because that retailer matters to them. Enthusiasts follow companies they already favour. Prior purchase, category interest and brand familiarity can create both recall and sales, which means aware and unaware groups may have differed before the campaign appeared. Heavy category buyers may simply create more opportunities for both purchase and recognition. Advertising could still have contributed. The simple comparison cannot tell us how much. Holdouts, experiments and matched starting behaviour ask a harder question: what changed among people who began from a similar position? Claimed recall and actual exposure deserve separation too, since memory itself is entangled with prior preference.

The danger arrives when measurement is commissioned as ceremonial proof. A flattering difference receives arrows, colours and a causal verb until the study appears to have isolated something it merely observed. The Rosser Reeves effect is decades old. The temptation keeps renewing because “buyers notice brands they buy” is commercially dull, while “our communication created valuable customers” fits beautifully into the conclusion prepared before the analysis.

Behavioural principles

Behavioural ideas at play in this post

Short, plain-English explanations of the principles behind this post, with links to related books and examples in the archive.