Hello,

I'm trying to determine best practice when it comes to the lookup function and report building.  Using the following example: Model object contains a identifier to that allows for a lookup to another dataset to get 10 additonal datapoints.

From a project performance/calculation times, which option is the best practice:

1. Create the lookups in the report(s).
2. Create the lookups in the dataset backing the model object.

Option 2 makes building future reports easier, but I'm really looking for guidance around performance and calculation times.

Thanks,

Jeff

Hi Jeff,

The general guideline I would recommend is doing the lookup at the data level, as it makes things a lot easier for you in terms of documentation, management, and future use.

An example of when I may use the !LOOKUP_COLUMNS function at the report level would be when I want to bring in columns to a report from a table that I don't have direct control over (Within your "Help" link, see: Technical Reference -> Hidden Apptio Tables, for examples).

Hope that helps!

-Michael Verkruyse

to add to Michael,

anything you do in a new column on the reporting surface is not cached long-term, even when precalculated. This is because as you apply slicers/groupings/etc, it has to recompute anyway. Since lookups are expensive, they are better in the data layer.

HOWEVER, really the right tool in my opinion [though sometimes there are reasons not to use it] isinference. If you inference link the dataset that you are doing the lookups to into your model object, than you'll be able to report directly on the columns from that dataset without doing lookups. Unless both datasets are backing model objets in different parts of the model this can often be the best option. If you have reference datasets, link them to your model object, and nothing else, and all of the columns will be available in reporting.