Transforming and coding ordinal variable based on relationships interaction between it and dependent variable

Question

I am trying to fit a linear regression with ordinal and nominal predictors for forecasting. For example, I want to forecast profitability based on whether it's a weekend/working day (but I also have other various predictors). I can dummy code this variable, however I want to make it more precise.

For instance, I can analyze the historical relationship between the day of week and profitability. Throughout the bunch of data, I see that on average from Mondays to Thursdays profit is almost the same, on Fridays it's x1.5, on Saturday it's x2, and on Sunday it's x3.

Can I use these findings to arrive at finer decoding so that this variable takes the following values (for demo purpose): Monday to Thursdays – 1, Friday – 1.5, Saturday – 2, Sunday – 3? Or is it wrong because there are other regressors that can moderate this relationship? Other concerns? If it is wrong how can I code this variable better provided that I have other continuous exogenous factors that are correlated to the variable in question – number of visitors, engagement, etc?

Thanks in advance!