When is this change happening?
On October 25, 2023, our v3 API endpoints will be redirected to our new count system. You will not need to make any changes to your APIs. However, you may see a slight change in your count.
What is changing for Entry data?
We are switching the source of the count data for Entry with a new system that is faster, more forgiving, and more accurate. This translates to a new database, with the same metrics and views in Atlas. Here are the main differences:
- Better drift correction: Our new drift correction algorithm has dramatic improvements. Drift correction normalizes throughout the day, rather than the steep drop you previously saw at the reset time.
- Better historical data: Changing sensor orientation, doorway locations, and creating new spaces from existing doorways will automatically backfill historical data with those settings. Previously, this resulted in data updating only from the time of change. The system will now retroactively apply all changes to the data.
What is changing for OA data?
- Better historical data: Making changes to your configuration will automatically backfill with those settings. Previously, this resulted in data updating only from the time of change. The system will now retroactively apply all changes to the data.
- 30-Second Sampling Rate: The old data pipeline sampled OA data at a 1-second rate. We now sample at 30 seconds in the new pipeline. This improved both performance and accuracy. You may see a slight difference in occupancy max and time used endpoints, but most metrics are unaffected at lower granularities.
What is drift?
”Drift” refers to the amount by which the count of a space in our system differs from the actual number of people in the space. There are two types of drift:
- Positive Drift: Occurs when exits are missed or extra entrances are counted by a sensor causing positive counts
- Negative Drift: Occurs when entrances are missed or extra exits are counted by a sensor causing negative counts
What is the difference in the drift correction?
We previously corrected drift with two methods: Daily resets and Max Dwell. Both of these involved inserting entrances and exits at a specified time to correct the count to 0.
Now we will normalize the data across the day, from the assumption that the count should be 0 at the daily reset time. When you query data, we adjust the slope of either the exits (in the case of positive drift) or the entrances (in the case of negative drift). This results in the same trend over the day, but without the steep drop-off at the reset time.
Before:
Visible drop in occupancy at the 4 a.m. reset time, with the count staying at the same high number from 6 p.m.
After:
Data is smoothed out naturally across the hours, taking into account that the count should be 0 at 4 a.m., but occupancy did not significantly change after 6 p.m.
Drift Correction Changes
- Drift correction is applied at query time. Changing the daily reset time will change the results for all historical data. You can adjust the daily reset time multiple times to find the setting that makes the most sense for your space.
- Daily resets should be set for when your space is completely empty, including cleaning crews and night staff.
- The daily reset time default is set to 00:00. You can adjust this at any time.
- Drift correction is only applied after the daily reset time. Please wait 24 hours to look at data for Entry.