This paper proposes a new definition for trade direction to capture the information effect in order-driven markets. To define the trade direction, the existing literature uses quote-matching rules to identify whether it is the buyer or the seller initiates or triggers a trade. The quote-matching rule is well suited for a quote-driven market, but not so for an order-driven market. In a continuous auction and order-driven market, orders that are submitted based on information but do not trigger a trade will not be counted using the quote-matching rule. Also, in a call auction order-driven market, one cannot identify who triggers the trade because the timing of order clearance does not depend on the arrival of any specific order. To deal with these issues, we propose to define the trade direction based on the sign of the net order arrived between two transactions, that is the relative size of the buy and sell orders. If the buy order arrived between two transactions is larger than the sell order, we say the trade is buyer dominated and define the trade direction to be positive. Based on the new definition, we develop ATT (Augmented Tick Test) rules to infer the trade direction. To examine the validity of this new trade direction, we apply our ATT rules to the transaction data observed in the Paris Bourse and the Taiwan Stock Exchange and find that ATT rules have information content. Furthermore, using the ATT rule and quote-matching rule together allows us to decompose the trade-related return into permanent and temporary components. The ATT rule is better at capturing the permanent component, while the quote-matching rule is better at capturing the temporary component.