More information is available from single source panels than is often realised. This includes descriptions of the category, of the brand, the associations between brand share and price, demographics, weight of viewing and recent viewing. Data can be aggregated into weeks and normal time series modelling compared with the disaggregate findings; the latter seem to be the more sensitive. Reasons for
... [Show full abstract] the brand choice at each purchase occasion can be studied by multivariate regression. These include the shopper's loyalty to the brand, its relative price, a trend term and recent advertising for the brand and for its competitors measured by adstock. Short-term advertising effects have been seen at two to 28 days half life for various brands; no effects have been found for some others. Competitors' adverse effects may be larger or smaller than ours. Diminishing returns to higher current advertising pressure can also be measured and are usually slight. A minority of occasions are under high pressure and most of these are for heavy viewers who are also affected by competitors' activities and often have untypical brand shares. Any bivariate relation between recent advertising exposure and brand choice is potentially affected by purchase/viewing bias which often occurs and by other confounding factors such as price. Such relationships can give misleading indications of advertising effects.