Regression Model - Solver Issue With 15+ Independent Variables

According to this: https://www.solver.com/standard-exce...mits-continued Solver is limited to 100 individual constraints. Your Solver model has 1400+ cells in the one constraint, and I think Solver is treating that as too many constraints. I think you need to examine your regression model and see if you can reduce the number of constraints somehow or obtain a more robust solver engine (either from Frontline or other source) that can handle that many constraints.

Ah! That makes sense. Since the constraints are applied to my forecast column itself, which contains a sample of 1,000+ menu items, should I just remove those constraints? The problem is that branded purchase intent is a percentage, so it would need to be between 0 and 1...any way to force this constraint without applying it to 1,000+ rows?

I deleted that constraint and let the solver model run and it converged just fine. I added a COUNTIFS(E6:E1449,">1",E6:E1449,"<0") function to see if any were outside of the 0 to 1 range. So the results of the forecast column seemed to naturally meet the constraint without explicitly declaring the constraint. Have you encountered scenarios where this constraint is violated? Can you identify why those cases exceeded the constraint? I think I would want to explore scenarios that caused it to violate those constraints and see if I could understand what causes the violation, then address it that way.

After adding that COUNTIFS() function, you could make the constraint (countifs function =0). However, Solver sometimes has trouble with counting functions because they are not continuous. That might be worth looking into.

Really? When I remove the constraints, the forecast column now only shows "0" and 1's," which is not realistic for a branded purchase intent score for menu items as they are typically a percentage between 0 and 1. Perhaps you could post a link to your version?

I did not save my version, though I recall that the format on column E was "integer (number with no decimals)" or something like that -- naturally causing numbers between 0 and 1 to display as either 0 or 1. You should be able to apply whatever number format you think is appropriate to better see the actual values returned to these cells.