Regression Forecasting Model

Hope someone can help me with some fairly favourable pointers on this subject.

I have historical data to draw from and wish to forecast possible turnover according to parameters I can use taken from these existing analogues.

My historical data consists of the performance of outlets before and after they were assigned an business investment, as below.


a) Old turnover
b) Business investment type a)
c) Business investment type b)
d) New turnover
e) Turnover uplift percentage
f) Old size of outlet
g) New size of outlet
h) Turnover by size of outlet
i) Value of customer market
j) Old percentage share of market
k) New percentage share of market

With a number of records to draw this information from, am I able to 'plug' this data into a template regression model to then make forecasts from?

When creating a new forecast and to place a figure against parameters d), e), h) and k) I would have available to me the other parameters in place, namely:
a), b), c), f), g), i) and j)


I have no previous experience of creating/adjusting a regression forecast model so really am after some in-depth help here.

Many thanks in advance.

Are there any forecasters out there than can lend some help on this post please?

We are here, but I'm not sure that even the best among us will be able to help you based solely on the information given.

The first thing I note is that you have 10-12 input parameters for this model. That is a lot of information to try to regress into a single model. In addition, you admit to having no experience with regression/forecasting, which means that I would anticipate that our task is significant.

With that said, I will suggest a couple of things to hopefully get you started in the right direction.

1) Arguably the most important part of a task like this has nothing to do with Excel. The first and possibly most important step is to come up with a function y=f(a,b,...,k,time) that makes sense for the type of data you are regressing. Your description sounds like a financial type model, and, being a scentist/engineer, the nuances of coming up with financial function tends to elude me. Forecasting implies that you are extrapolating into the unknown, which makes this step even more important, because how well a model will extrapolate depends greatly on how reasonable the model is.
2) Once you have decided on the form of the function, Excel provides a couple of different tools for doing the regression. If the function is "linear" (in the linear algebra sense, if that even means anything to you), then you can use the LINEST function to do a least squares regression. If the function isn't linear, then you can try using Solver to do the regression.

Are you using a scatter plot and building a linear regression based on the line of best fit? Something else?

If linear regression based on a line of best fit, your generic formula is y = a + bx, meaning you're working with one variable. In this case, you are comparing the effect of that variable with respect to some constant.

If you're working with another type of regression model, you have to account for omitted variable bias and dependent/independent variables (see Gauss-Markov theorem). In sum, you should be isolating the effects of each variable with respect to the final result (i.e possible turnover). This is the beauty of this type of analysis.