# timeGroup function group data on Thursday **URL:** \ **Category:** Grafana\ **Created:** [June 26, 2020, 10:09am UTC](https://community.grafana.com/t/timegroup-function-group-data-on-thursday/32750 "2020-06-26T10:09:35Z")\ **Posts on this page:** 1\ **Showing post:** 3
**Author:** ![jerimiahmack](https://avatars.discourse-cdn.com/v4/letter/j/958977/32.png) [@jerimiahmack](https://community.grafana.com/u/jerimiahmack)\ **Post date:** [March 25, 2021, 7:14pm UTC](https://community.grafana.com/t/timegroup-function-group-data-on-thursday/32750/3 "2021-03-25T19:14:37Z")
It is because of the way that the $\_\_timeGroup() function calculates bins of 7 days. In the case of MS SQL Server, it applies the following: FLOOR(DATEDIFF(second, ‘1970-01-01’, \)/604800)\*604800 1970-01-01 is considered “0”. It also happened to be a Thursday. 604800 is the number of seconds in a week (60 x 60 x 24 x 7). The DATEDIFF()/604800 portion calculates the number of elapsed weeks, including partial week, since 1970-01-01, with second precision. The FLOOR part drops any fractional week from weeks elapsed, then calculates the number of seconds corresponding to the number of **whole** weeks elapsed, which returns a ‘7-day bin starting on a Thursday’. $\_\_timeGroup() really should allow a referential offset, but doesn’t. Rather than using $\_\_timeGroup() you may wish to affect you own FLOOR logic based upon 1970-01-01 + Xday offset to reset your “0” point to the following desired start day. For example: 1970-01-04 (Sunday) or 1970-01-05 (Monday). You could use a dashboard variable to map multiple bin widths (hour (3600sec), day (86400sec), week (604800sec), month, etc.) to allow for quick precision changes. --- _[View the full topic](https://community.grafana.com/t/timegroup-function-group-data-on-thursday/32750)._