Help me get started setting up an elementary problem for use with a monte carlo sim

[dude]

I want to further understand monte carlo simulations. I have a business and I want to correctly set up a problem to find the likelihood of future sales of each product line. I have historical data and access to anything I need. I have ranges (high and low) which we believe to be possible this year.

I need help narrowing down a clear objective and then getting the problem set up in terms of the basics. To be clear, I am not looking for excel help to start with. I want to understand how/what/why/when to set these up and use them.

So where do we start?

Thank you in advance. I just want a better understanding so I can further develop my skills in this area for future projects.

[Dason]

I'm not exactly sure what you're looking for? Can you explain a little bit more?

[dude]

I am wanting to run a monte carlo situation on a real life situation. I wanted to use an example I can relate too. Sales forecasting.

Make more sense?

[dude]

Or, can anyone reccomendi a good book with plenty of examples? That may be better at this point.

[Outlier]

Start with a histogram of past performance using 5 to 20 bins.
You can use Excel to multiply/divide/add/subtract these histograms to get your prediction, but it will be less labor intensive and tedious to program this in BASIC, Pascal or some other language, using nested loops.
Even if the histogram looks like a normal distribution the answers, in general, will not.

[dude]

Thank you. I can get the bins but I am unsure what you mean by subtract/multiply/etc. the histograms to get the prediction. I am going to have to use Excel since I am not proficient in any programming language.

---

In the mean time, I did a quick sales prediction based on my knowledge of product groups. I just dialed in a Min and Max and then introduced a random() to get a number between the Min and Max and ran it 2000 times. Then took the average of the row sums.

Any smart way I can introduce an expense into the above so I can better reflect as sales go up so does cost?

[Outlier]

The min/max probably won't give you probabilities of intermediate values. Post some simple examples of your inputs and desired output.

[dude]

Min: $10,000
Max: $20,000


=ROUND(RAND()*(D$8-D$7)+D$7,2)

D7 = Min
D8 = Max

I have seven columns this way for expenses. Obviously, this hits very close to the average/median due to the distribution the Random() function spits out.

I am working on a way to tie in expenses ad dependencies.

I have been advised to look at correlation with one cost vs. expenses to find r and r^2 will tell you what % of expenses will be attributable to that cost. Then you can use multiple regression for all the costs vs. expense.

Thoughts?