What an absolute joke your system is!!! Today, your system went backwards in working out what was what when. If you are not going to supply consistent data in regards to how you collect it, get rid of the thing, it is farcical.
Mmh, not entirely sure what you mean? 
The values are the median, the middle value of a sample.
For my site, that has not changed, e.g. for flights the middle value of the last 7 days is 3128, and that is shown on the nearby sites table.
3065 2.
3017. 1.
3272 7.
3237 6.
3173 5.
3128 4.
3079 3.
2 Likes
Oh I under stand the medians etc. Example yesterday 4th they used those figures from 3rd July. Today 5th they have gone back to the 1st July for the figures !!! The prosecution rests your honour…
It’s the 7 day median, with the extra twist that one of those days is the current day. The current day always starts at 0, and increases through the day. That means that the value for the current day is going to be less than the 7 day median for most of the day. It is probably going to be the lowest value of the 7 days for most of the day. As the day progresses, if it is a high volume day, the value for the current day will exceed some of the lower days. It may cross over the median value, causing the median to flip to some other day. The median will continue to track up with the current day’s value, until it stops at the next highest day in the list. But then at midnight UTC the 7th day fall out of the calc, and a new day starts with a zero counter, and a different day is likely to be come the median.
The confusion comes from using the current (incomplete) day in the median calc. If it was based on the previous 7 completed days, ignoring the the current day, then the median value would only change at most once per 24 hours. The value would have more stability, but it would never reflect the current day’s activity until after the end of the day. As it is now, a high-traffic day moves the median up before the end of the day.
The value is the median of 6 complete days, plus the current incomplete day.
Maybe it has something to do with different time zones. On your data feed it says 07.07.18, the daily data table still collects data for the 6th, because that value is not in local time but utc, and your time zone is utc+10.
If we were talking about a 10 hour difference, yes, but not going backwards 3 days, kind of doesn’t fit the UTC / local time difference. Now the current figures are using the completed Friday amounts for the calculations………… We will see what day the next calculation uses 
Yes, I thought that might explain a change during the day. The moment the utc day is completed, one day drops out of the 7 days used to calculate the median.
Why shouldn‘t the median jump back three days? If that is the middle or 4th day of the 7 after they have been ordered by size?