As long as all the data will fit in memory, you can store a result set from a transform in the containing job with the Copy rows to result
step. Then in subsequent transforms you get the results from the job with Get rows from result
.
The 'result' they're referring to is just a big buffer that's stored in the job, so it will be available to any transform that is contained by that job. It's similar to the files 'result' which records all files touched by a job.
If your data set is too large to keep in memory, you'll need to stage the data somewhere. Either in a database table, or a file somewhere. You might look into using the Serialize to file
step for that.