Am I missing something here but how can you have an average price of £1.03 with a SD of £40.25 - This would mean delivery on some items is £-39.22?
I am working on an analysis of new delivery charges and and comparing existing pricing with a proposed new pricing.
I have 84k orders which I have calculated the average and standard deviation of difference in pricing between current/proposed pricing.
avg = =£1.03
stdDev = £40.25
80k of my order fall within 1 stdDev and in terms of value its £150k positive.
Is a stdDev of 40.25 good or bad? high or low? should it be closer to the avg in order to be more useful as info??
I wouldn't say there is such a thing such as a 'good' or 'bad' standard deviation.
A high standard deviation usually means that the value that you can obtain (here the charges) are not easily predictable.
It's like I tell you that I'll give you 100 dollars with a standard deviation of 50 dollars each day.
This means that one day, you might get 0 dollars (in fact -50 dollars) and on other days, 250 dollars. Those are not so close to the average, 100 dollars.
Whereas if I tell you I'll give you 80 dollars with a standar deviation of 1 dollar, you can get 77 dollars or 83 dollars so you can predict the value that you'll get with some accuracy.
Of course, I'm taking the standard deviations from a normal distribution where 3 standard deviations fall within 99.72% of all possible outcomes.
I am trying to figure out how much we stand to gain or lose with a new delivery pricing model. I have 84k orders which i have applied the existing and proposed pricing to. Its the difference between these two charges that i'm working with and the price often shows a loss so the stdDev goes into minus figures.
94.68% of my data fall in the first stdDev (33.53% negative, 61.16% positive)
3.82% fall in the second stdDev (1.30% negative, 2.52% positive)
1.46% fall in the third stdDev (1.46% negative, 0.03% positive)
now i can see that 94.68% of my differences in charges are between -41.28 and 39.22 but how does that help me?? the difference of £40(ish) is significant so is the result that under the proposed charging the recovery is, on the whole, very different to existing charges? is that all i can assume?
I wouldn't say it is very different to existing charges, but it can be very different.
Taking my previous example as a sort of 'bet'.
Some people would rather bet on more money even if there is greater risks!
Others will tend to bet on smaller sums which they can better predict and perhaps win on the long term.
From those figures that you gave, I would say that you are like the person who is taking much risk where you can suddenly win big, or lose a heap at once.
This all concerns probability but in reality, it doesn't necessarily be that way.