Since the advent of standard national accounts data over 60 years ago, economists have traditionally relied on monthly or quarterly data supplied by central statistical agencies for macroeconomic modelling and forecasting. However, technological advances of the past several years have resulted in new high-frequency data sources that could potentially provide more accurate and timely information on the current level of economic activity. In this paper we explore the usefulness of electronic transactions as real-time indicators of economic activity, using Canadian debit card data as an example. These data have the advantages of daily availability and the high market penetration of debit cards. We find that (i) household transactions vary greatly according to the day of the week, peaking every Friday and falling every Sunday; (ii) debit card data can help lower consensus forecast errors for GDP and consumption (especially non-durable) growth; (iii) debit card transactions are correlated with Statistics Canada's revisions to GDP; (iv) high-frequency analyses of transactions around extreme events are possible, and in particular we are able to analyze expenditure patterns around the September 11 terrorist attacks and the August 2003 electrical blackout.