A tick database is a specialised database designed to hold market data at its finest level of detail, with each record representing one tick: a trade, a quote change or an order book event, together with a precise timestamp. A single busy market can generate billions of such records a day, far more than a general-purpose relational database handles efficiently. Tick databases therefore store data by column, order it by time, compress it heavily and often keep recent data in memory. They provide query functions suited to a time series, such as joining each trade to the quote prevailing at that instant.
Trading firms use tick databases to run backtesting on realistic data, to perform transaction cost analysis, to monitor trading for abuse and to answer regulators' requests to reconstruct past markets. The main costs are storage, licensing of the data itself and the effort of cleaning errors, duplicate records and inconsistent clocks.