How do I split a proportion of a future calculated value in two dimensions

Trying to get my head around Causal - loving it so far! Now I’m trying to build a model that forecasts the output of product by country over time according to an overall growth rate (say 2%) but where the output is split into sub-categories (say % low, medium or high quality). See below…

So, Argentina might have an overall annual growth rate of 5%, but 12% of that additional output will be low quality, 25% medium and 63% high (must add up to 100). My model is monthly over 5 years and I need the total volume of production by country by quality each month.
Is that possible?
Thanks for your help.

Hey @Tim_Wilson,

Yes, you can use nested categories to accomplish this!

Here’s a demo model that you can clone to get you started – all you’ll need to do is create a new category called Quality and apply that to the Production Quality variable.

Thanks @Tim_Nowacki! This doesn’t accomplish quite what I was looking for. I was hoping that I could get Causal to produce the output by month by country by output type and be able to ‘roll up’ / drill down all of those. So output in Month (current) for Argentina is <annual growth factor (%)> higher than in Month (previous) and X% of that output is (low) (from <country production profile - low (%>), Y% is (med) (from <country production profile - med(%>) and Z% is (high)
I don’t think this is possible. I have just read the documentation in a bit more detail and spotted this…

  • which I think answers my question… :disappointed:

Hi @Tim_Wilson - yes, you can accomplish what you have in mind. We don’t encourage many layers of “nested” categories but you can use them in this case. Please see the demo model here, which I’ve tweaked from the version that I sent yesterday.

  • Production Quality Profile lists the production profile over time (this can be time-varying)

  • Total Output shows the total output over time by country. I hardcoded random numbers here.

  • Total Output by Country and Quality Type is the product of the two variables above and should accomplish what you’re trying to do!

Ah! I get it now! Many thanks for your help @Tim_Nowacki


Hi Tim - I note you said you “…don’t encourage many layers of “nested” categories…” - is that because of application performance…? I now have 4 layers - one of which is cohort (on a 5 year monthly model) and it’s becoming unwieldly. What is best practice here - to get the power of nested categories…?

Hey @Tim_Wilson,

Jumping in here. Yes, one of the issues with multiple layers of categories, especially cohorts over longer time periods, is model size and therefore performance. We do have enterprise level features that significantly improve performance and allow for much larger models so if performance is becoming an issue I would suggest considering upgrading :slight_smile:.