TM1 Forum

Hello

We have a presentation from a project of one of our customers to many others customers in a meeting. We aren't allowed to show the real data, and therefor we want to change the data in cubes to random values. Unload and reload with a factor is an possiblity to do so, but takes some time on all 36 cubes.

Are there any other ways to handle this easier?
In my mind i think there was an option "falsify data" by rightclick on a cube, but was it just my imagination or is it gone?

Best regards
Jani

Write a set of TIs to read the values, multiply them by (.9 +rand*.2) and write them back.

Just what every reporting or planning application needs where data integrity is vital. An automated function to falsify data!

I do recall that at some point IBM support did offer a command line utility that operated on the file system level on .cub files (while server down) which did 2 things
- replace cell values with random values
- thin out data by deleting every nth record
The intent being for packaging data sensitive models for support where there is stability other issue needing support.

I think this would do just what you need, if you can get a copy from IBM support if it still exists or is maintained.

Otherwise use spreading or do a DYI TI process to loop through cubes and randomize data as per David's suggestion.

Both these approaches will effectively randomize data but often the most sensitive data is the identity of customers, products, employees, etc. If these need to be anonymized there's no easy way around it.

Interesting to hear of a utility to do the randomising.
How would spreading randomise? It can change data but not randomly, I would think.

Yeah spreading wouldn't randomize but a few spreads here and there and you've effectively masked or made unrecognizable from the original.

David Usherwood wrote:Write a set of TIs to read the values, multiply them by (.9 +rand*.2) and write them back.

I did something very similar recently but also did *-1 if RAND < 0.5. This confused things even more