The amount of effort put into getting reference data correct at big banks is huge, but the volume of data is insane and often there are no good places to verify

The amount of effort put into getting reference data correct at big banks is huge, but the volume of data is insane and often there are no good places to verify it. I had similar experiences when supporting the ref data platform. On one system holding ETF and Index data we had several thousand products each holding between 10 and a thousand components (eg DJI through Russel 3K). On each product and each component of it there are multiple major fields that need to be correct (price, quantity, multipliers, etc). Any mistake throws off the risk systems and trading in those products (eg basket trading an index, delta hedging, etc) Millions of data points, and single data errors caused big trading errors. The data itself could be wrong from the actual providers too, taking debugging into a whole extra level of complexity.