Demand Modelling Tool - data ranges for seasonality
I’ve been having a look at the modelling tool. I’ve got weekly data from 01/04/2019 to today but the tool says there is “insufficient data to calculate seasonality”. Do you know if this is about weekly volume?
It might be because our model line is just a straight line with a “bump” in it, depending on how long I say it takes us to recover. Should the modelling line be spiky like our true data?
12 Views
Because 2020 was a weird year, we need data back to 01/01/2019 to calculate “normal” seasonality usefully. We can’t use 2020 to calculate normal seasonality. Without that, you’re seeing the model’s calculation of what happens before, during and after lockdown starts and ends which will be a series of flat “average” lines. The seasonality then adjust those flat lines for historic ebbs and flows in your data.
The obvious logical conclusion from this is that the model is going to get less useful the further into the future we get – how long are we going to treat 2019 as a valid reference point for normal seasonal trends? This is why I’m pushing for the tool to move away from this kind of modelling and look instead at system flows – of the referrals we get, how many end up as CLA, and how long do they take, etc. It’s not enough that it would be useful to do the kind of modelling we are currently doing – it also has to actually work in the longer term!
Just to note if you can’t get annex a reports to go beyond 01/04/2019 you can get weekly referral counts from any other source and paste them directly into the summary data tab instead of having the tool convert from annex a.