Looking for a simpler solution

We have a project here that we are working on, that has a report summary that we are trying to determine the best way to get the reports, based on speed and sever load. Here is the scenario:

A client can go in and adjust a value used in a calculation, say he goes in several times to update it:

'
Now, for the reports, we need to find the average value for the month of May. The prototype I'm building off of went and took all the average for all entries for that month ((9,000 + 5,000 + 6,000 + 5,500)/4) = 6,375.

However, these values are for daily activity of a business, so this method doesn't take into consideration that the spike of 9,000 only lasted 3 days and the initial value at the begining of the month (5,000 set back on April 29th). To get proper value we are looking for, we would need to do:

'Here we get the correct average for the month, about $500 less.

Now, I'm seeing the easiest way program wise to add this to the program is, whenever you add a record for a day, it goes and adds a value for each day since the last insert (i.e.. on May 21st, it would loop through 8 times and insert a record for the 14th - 20th (with the value from the 13th) and then the new value of on the 21st. Then we can call the database and just get an average where month = May.

However, we are wondering if there is a simpler method to do this where is calculates in a value for missing days as well. Additionally we are concerned over if we did do our own function to do the caclulate missing values between the entry days, would the speed of such a function be considerable. (Keep in mind the reports are eventually going to be able to span over several months, and hopefully if things go well, years).

Thank you in advance for any suggestions on making this problem more simplified and optimized.

-Greg

PS. The main downfall I see to my easiest solution is, there may be clients who may not do any updates for a few months, and for 3 months, that would be running about 90 INSERT statements...)