Hi. I am implementing volume anomalies test on a model where I want to check data per date in a timeframe of -45 days from now and any spike or drop within that frame should be flagged. I have set days_back to 45, but it seems it actually goes 45 weeks back when inspecting the result in the report as it displays january 18 as the start date. Am I missing something? this is my config:

models:

- name: my_model

description: my model

config:

elementary:

timestamp_column: as_of_date

tests:

- elementary.volume_anomalies:

anomaly_sensitivity: 2

seasonality: day_of_week

days_back: 45

backfill_days: 45

time_bucket:

period: day

count: 1

config:

severity: warn

Hey @Anders Enger I'd say it's because of the seasonality setting. Is it defined intentionally ?

yes, because I only have data on weekdays mon-fri, so I want it to compare saturdays with saturdays where it is normal to have row count 0, so weeked 0's should not be considered anomaly. I see now the seasonality will multiply days back with 7, but couldn't quite get it to add up.

what do you mean by get it to add up ?

If I set it to 7 days the start date becomes Nov 8 which is 12 days back, but if I set it to 45 days the start dates becomes Jan 18 which is 306, while the documentation states days back will be multiplied by 7, which is not the case in either example

@Anders Enger Can you send over the test config with the 7 days, and a screenshot of where you see the starting date ?