Somebody ask me to migrate an excel file (10000 rows, 600 columns) to an existing sql server 2000 database. Is not as simple as simply upload the file to the db, but I will need to put data from a single sheet to several tables, format it and the result has to be compatible with the sofware using the db. They don't have documentation for the db but at least the fields names are descriptive). How much would this cost in usa?
A lot. It'll depend on how many databases, and how complex the data is. This is the type of work my father does for banks, when a bank merges with another bank they choose one bank as the main database, and then they have to spend a great deal of money and time making it so that first there's a way for administrative users to access all accounts, and second migrate the data over or build connectors so no one needs to know or care which database the customer is in.
You typically have to run the data against a test database, then test the operation of the software to make sure nothing broke. Without proper documentation your job is significantly harder, strange character sets, odd usage (phone numbers with or without hyphens, parens, etc), whether name prefixes and suffixes are in a separate field, etc, etc, etc all make this a much more difficult task than you might think in the beginning.
Then it sounds like you have a normalized database, and data which isn't normalized to enter. This means that you might have a dozen lines that relate to the same entity, and should go into one table as one row, and another table as 12 rows, and a third table as a few different rows, all related to each other. But what do you do when it's clear that these twelve rows are the same entity, but they have different data in one column that should be the same (say, customer address). Do you create two customers and hope it doesn't cause problems down the road, or do you choose one and hope for the best, or do you have to hire someone to track these entities down and "fix" the data you're entering?
Worst case scenario you hire data entry operators to print and re-enter the data, but at 180 items/hour, $20 per hour, and 6,000,000 items to enter it's just under $700,000. Further this doesn't fix the normalization problems if the data isn't perfect.
It's a one man, one week job if there are no problems, if the data is perfect, and if the massaging required can all be programmed in (changing periods to commas, for instance, if the number systems are different).
It's months and a team of 10 people if the data going in is bad and the software requirements are unknown.
