How should I interpret to negative constant in multiple linear regression in python?

Question

I have a dataset. A part of the dataset is as shown below:

sales_date  net_sales_my_firm   net_sales_others    pro_unit_my_firm    pro_unit_others
1.02.2021   101089              710337              9869                67885
1.03.2021   104747              598684              9084                79405
1.04.2021   92027               623285              8025                122489
1.05.2021   85796               463898              7541                63562
1.06.2021   112804              621633              10553               83586
1.07.2021   89326               484894              7832                61799
1.08.2021   85406               524195              7551                75599
1.09.2021   131388              686136              12144               87755

net_sales_my firm: Net sales of my company net_sales_others: Net sales of competitors pro_unit_my_firm: promotion sales of my company pro_unit_others: bonus sales of competitors

What I want to do is find the effect of promotional sales on net sales. For this, I used the multiple regression code I specified below (in python).

Y = df.net_sales_my_firm
X = df[['pro_unit_my_firm','pro_unit_others']]
X = sm.add_constant(X)
model = sm.OLS(Y, X)
results = model.fit()

Results summary is:

Results summary:

print(results.params)
const                -14896.842089
pro_unit_my_firm      4.163607
pro_unit_others       0.806564

I interpreted this result as follows: If you promote 1 unit, you increase your sales by 5 units. But; What does a negative constant value mean? Is it normal? Did I set up a wrong model?

I'm also sharing the scatterplot of the yield to help:

scatter_plot_data

Answer 1

I would have interpreted it as follows:

"If you promote by 0 units, and competitors promote by 0 units you will have negative net sales".

Maybe this isn't completely unrealistic. If you don't promote your product at all maybe you won't sell much, and so your net sales might be negative (depending exactly what net sales means).

This isn't what is implied by your plot where you have several points at 0 pro_unit_my_firm but of course it's not clear what the values of pro_unit_others at that point are.

To reassure myself that the fit 'makes sense' I'd want to inspect the values of pro_unit_others for the points where pro_unit_my_firm are zero.