Forum Discussion

hlorosc's avatar
hlorosc
Cloud Apps
11-22-2021

Tips on Managing Formulas / Getting Consistent Calculations

Question for y'all: How do you maintain consistent calculations of key metrics? 

Custom columns in the data model won't work b/c I need aggregations that can be easily filtered (e.g. can't add activation % to my base table b/c my stakeholders want to know what it is for different industries etc).

Formulas work great, but there isn't an easy way to lock them so I know the other analysts use the exact same ones. Plus, if I change the underlying data model or if I decide to update the formula, I have to start over.

Would love to know how y'all deal with this? In case it matters: we have a Sisense hosted deployment and are leveraging live models

Look forward to learning from what's worked in practice!

Data Models

4 Replies

Replies have been turned off for this discussion
  • irismaessen's avatar
    irismaessen
    Data Pipeline
    11-24-2021

    We're in a situation where our dashboard design is done by a very small number of people and dashboards can be audited before going live so we haven't had to be creative in this regard.

    What I've found happening often is that widgets get copied from already audited dashboards and then adapted further, using the authorised formulas. But this is not immune to changes in the authorised formula or the data model. Once a widget is copied, it is its own thing.

    Starred/Names formulas are I think the official way to share predefined formulas for metrics with other analysts. What you cannot do, however, is force people to use them, or stop defining other named formulas themselves. And indeed, if you change the datamodel or the formula, you still need to start over because I think old widgets/dashboards will keep using the old formula. I'd love to be wrong on that last one.

    I think documentation here is important no matter what you do. Not only does that allow you to explain exactly what each metric calculates, it also lets you explain how that metric is arrived at and what the limitations are (for instance, filter-wise). People should want to be using the same, relatively simple, formulas because that means they can properly compare numbers.