2 of 2 people found this helpful
Yes, you can load last 30 days data, as long as the data is loaded in chronological order, otherwise out of order data will be thrown out.
However we're fixing some intermittent aggregation service issue and plan to push that to production next week. I'd recommend to wait for that to be in place, as it will ensure the baseline is generated properly.
Could you also provide details on how the Baseline are calculated?
1 of 1 people found this helpful
It starts with first few hours data, discard the outliers to come up with baseline, then improves the behavior overtime. It considers hour of the day, so if say 9am-4pm is your peak, baseline will adapt to it.
The baseline fix for sporadic data issue has been in production for two days and looking good. Brad can see his baseline now. You can now give it a try with the historic data.
one more question, now that you mention "has been in production for two days":
Is there any link/page in which I can see the current "version" of the product and the history of applied improvement?
To loop back, we looked into the data, with current logic we need at least 24 hours of continuous data to generate baseline, there are couple hours without any data, baseline will not be generated, as is with this particular case. So ensure there's a 24 hour window where you keep sending data (at least one in an hour). We're also looking at enhancing the baseline logic to be more forgiving - will generate baseline even with period of time without data, albeit less accurate. With current backlog, expecting this to be in place in a month or so.