Testing when one year's sales volumes vary significantly from this time last year

Hi all,

I have a puzzle regarding sales volumes and how to know whether, 6-months into the year, for example, my current volume is on track for budget or not. E.g. it might be -5% off where I need to be to be matching a profile based on last years figures but scaled to hit my target for this year - but when should I react to this as anything but "sometimes I'm up, sometimes I'm down" and when should I react and say "it is likely that I am actually going to miss my budget by the end of the year".

Here's a little more detail.

1.) Sales run for 10 months of the year
2.) Final position volumes have been growing at over 10% per year for several years, so seeing that I'm up versus this time last year isn't surprising - so we set a budget - a target for this year's final position
3.) then, so that I can track sales on a daily basis, I smooth my target/budget for this year across the 10 months of trading according to a profile of sales taking the last 3 years' performance. e.g. if on average I have sold 10% of my final position by the end of the first week, that's my marker, then maybe I've sold 50% by the end of the 2nd month and things just slow down for the remaining period with a flurry in month 10 as people rush to buy (we are selling tickets for a conference). So - the sales are by no means even across the year.
4.) this gives me a model or expectation for how my full year target/budget spreads out over the year, day-by-day.
5.) I track our actual sales position vs. this profile each day and report the % variance. Of course in the early days we can wildly above or below the expectation as a customer who places a big order comes in a day or two earlier or later than last year (and therefore a day or two earlier or later than the model assumes they will this year)

So - the exam question is how do I meaningfully interpret the variance vs. expectation on a given day? I presume I need some sort of population if I am to do any statistical test? So do I need to look at today's figure of say "we are at 37% of the final target figure and today is 70 days out from the event, but in the last 5 years we have never been below 40% of the final figure.... so we are in trouble"

OR - is there any way I can simply make a meaningful comparison between actual performance and the model performance? I guessed not since there is no assumed variance in an underlying population, right?

FYI - the profile across the years do look quite like each other in terms of shape, but begin to vary the further back you go, since the world changes and our customers and businesses evolve...

any thoughts on this - either resulting in sensible heuristics or indeed actual statistical tests would be greatly appreciated.