Excel Solver Limitations: What It Can and Cannot Do

How-to basics · Updated November 21, 2026

Excel Solver is best suited to small, well-structured optimization models: the standard Solver add-in can adjust at most 200 variable cells, and its algorithms do not guarantee a global optimum for every model. Its results also depend on choosing a compatible solving method, setting up formulas and constraints correctly, and checking the solution status.

What Excel Solver does

Solver finds values for selected worksheet cells that maximize, minimize, or reach a target value in a formula cell while satisfying specified constraints. For example, a business might choose how many units of several products to make in order to maximize profit without exceeding available labor or materials. The model consists of three parts: an objective cell, changing cells, and constraints.

In Microsoft 365 for Windows, open Solver from Data > Solver. If it is missing, enable it through File > Options > Add-ins; select Excel Add-ins in the Manage box, choose Go, and check Solver Add-in. Solver is a desktop Excel add-in, not a feature you can run in Excel for the web. Excel for Mac also offers Solver as an add-in, though its location and interface differ slightly by release; look under Tools > Excel Add-ins to enable it.

Solver changes the values in the cells you identify and recalculates the workbook to evaluate the objective and constraints. It does not decide whether the business assumptions are sensible, whether the input data is current, or whether the result is practical. Those remain the model builder’s responsibility.

The hard size limit

The standard Excel Solver add-in is limited to 200 changing variable cells. A variable cell is one Solver is allowed to adjust, not every cell in the workbook. A model with hundreds of thousands of rows of source data can still have only a few variables; conversely, a model that assigns a separate decision to each of 500 products or time periods exceeds the standard add-in’s limit.

This ceiling can force a modeler to simplify or aggregate a problem. For instance, instead of choosing a quantity separately for every store, product, and day, it may be possible to group similar stores or use weekly totals. But aggregation can hide important differences, so reducing the number of cells is not automatically a sound solution. If the real decisions cannot be represented accurately within 200 variables, consider a solver designed for larger models.

There is no single useful limit on workbook size that tells you whether a model will solve quickly. Formula complexity, the number and type of constraints, integer decisions, and the solving method all affect run time. Even a model below 200 variables may be slow or difficult; the variable limit is a capacity boundary, not a speed or accuracy guarantee.

Solver methods and their trade-offs

Solver provides three principal methods, and the choice matters because each makes different assumptions about the model:

  • Simplex LP is for linear programming. The objective and constraint formulas must be linear in the changing cells. For a correctly specified feasible, bounded linear model, it can find an optimal solution. Integer or binary restrictions add complexity and may make the search much slower.
  • GRG Nonlinear is for smooth nonlinear models. It uses a local search, so the answer can depend on the starting values and may be a local optimum rather than the best solution anywhere in the feasible region. Trying different starting points can reveal that results vary, but does not prove that the global best has been found.
  • Evolutionary is intended for models with discontinuities or other difficult, nonsmooth relationships, including some models involving choices such as rounding or conditional logic. It searches heuristically, so it may take longer and does not generally certify that no better solution exists.

Choosing a method that does not fit the formulas can produce an error, a poor result, or a misleading impression that the model is impossible to solve. Linear models are often easier to diagnose and solve than nonlinear models with many local optima. Integer requirements, such as making a number of trucks whole, can increase the difficulty substantially even when the underlying formulas are linear.

Model types that are difficult or unsuitable

Solver is not a general-purpose answer engine. It needs a clear numerical objective and relationships that Excel can calculate. It is a poor fit when the true problem depends on extensive branching logic, a huge number of yes-or-no decisions, simulation, complex scheduling, or data that changes unpredictably during a run. You can sometimes express part of such a problem in a worksheet, but the resulting model may be fragile or too demanding for the standard add-in.

Nonlinear models can have multiple feasible regions or local optima. For example, a formula whose output rises and falls as a decision variable changes may have several peaks. GRG can settle at one peak based on its starting values; another starting point may lead to a different answer. Evolutionary search can explore a wider range, but a plausible-looking result is not the same as a mathematical proof of global optimality.

Discrete choices also need careful thought. A model may need a decision to be an integer, binary, or otherwise restricted. Adding such constraints can turn a relatively manageable continuous problem into a combinatorial search. A model with many binary decisions can be particularly challenging because the number of possible combinations grows rapidly. Solver may stop at a time or iteration limit before it can establish the best result.

A small example: a production plan

Suppose a workshop makes products A and B. Put the quantities in B2 and B3, respectively. Put their profit per unit, 12 and 9, in C2:C3; labor required per unit, 2 and 1, in D2:D3. Put total profit in B4 and labor used in B5. The formulas, with those cell assignments, are:

  • With quantities in B2:B3 and unit profits in C2:C3, enter =SUMPRODUCT(B2:B3,C2:C3) in B4. This calculates total profit.
  • With quantities in B2:B3 and labor per unit in D2:D3, enter =SUMPRODUCT(B2:B3,D2:D3) in B5. This calculates total labor used.

In Solver, set B4 as the objective and choose Max. Set B2:B3 as the changing cells. Add constraints B5<=100, B2<=30, B3<=50, and B2:B3>=0. Use Simplex LP because both formulas and constraints are linear. With these illustrative values, the best continuous solution is 25 units of A and 50 of B: labor use is 25×2 + 50×1 = 100, and profit is 25×12 + 50×9 = 750.

This example is small enough that the add-in’s variable limit is irrelevant. It illustrates a different limitation: the answer is only as meaningful as the assumptions. If products must be made in whole units, add integer restrictions; if there are setup costs, production minimums, or other nonlinear effects, the model and method may need to change. A solution that ignores a real constraint can be mathematically valid for the spreadsheet but useless for the actual decision.

Why Solver can stop or give an unexpected answer

A Solver message is part of the result, not a detail to dismiss. Depending on the model and settings, Solver can report that it found a solution, reached a time or iteration limit, could not find a feasible solution, or could not establish an optimum. Reaching a limit does not necessarily mean there is no solution; it can mean the search was incomplete. Likewise, a feasible answer means the constraints were met within Solver’s tolerances, not necessarily that the model is globally optimal.

Check the worksheet after a run. Confirm that the changing cells contain sensible values, that all relevant constraints are included, and that the objective formula points to the intended outputs. Check whether Solver retained the solution or restored the original values, and review the displayed status. For models with integer decisions, verify that the returned values obey the intended whole-number or binary restrictions.

Solver works with numerical calculations and tolerances, so tiny differences can arise around a constraint boundary. A displayed value rounded to two decimal places may conceal a more precise underlying value. If a constraint must be met with a safety margin, model that margin explicitly rather than relying on how a cell is formatted. Also check that formulas do not refer to unintended cells, omit a cost, or use a unit inconsistently.

Reporting and sensitivity analysis

Solver’s reports are not equally available or informative for every model. Sensitivity analysis is most useful for appropriate linear models solved with Simplex LP; it is not a general report that explains every nonlinear or integer solution. When available, a sensitivity report can show how certain changes to coefficients or constraint limits affect the solution under the report’s assumptions. It should not be treated as a forecast of what happens after large changes or after the model’s structure changes.

For an integer model or a nonlinear model, inspect the chosen values and test meaningful scenarios directly. Change an input within a realistic range, solve again, and note whether the decision changes sharply. That kind of scenario check can expose a solution that depends on a fragile assumption, but it does not replace a specialized statistical or optimization analysis when one is required.

When to use another tool

Excel Solver remains useful for teaching optimization, prototyping, and modest models that fit naturally into a workbook. If the model exceeds 200 changing cells, takes too long, needs stronger diagnostics, or requires a reliable optimality guarantee for a difficult problem, compare alternatives. Commercial Solver products offer expanded capabilities, while tools such as OpenSolver or dedicated optimization libraries can support different model sizes and algorithms. Their exact limits depend on the product, solver engine, and problem type.

A specialized tool is not automatically better for every task. Moving a model out of Excel can require translating formulas, validating the translation, and maintaining code or a separate data pipeline. For a modest linear model, Excel may be easier for colleagues to inspect. For a large mixed-integer scheduling model, a dedicated solver may be more appropriate. Compare the actual requirements—variable count, integer decisions, nonlinear formulas, run time, reporting, and audit needs—rather than choosing on brand name alone.

Frequently Asked Questions

What is the maximum number of changing cells in Excel Solver?
The standard Excel Solver add-in can adjust up to 200 variable cells. The limit applies to the changing cells in the model, not to the total number of cells or rows in the worksheet.

Can Excel Solver find a global optimum?
For a correctly formulated linear model solved with Simplex LP, Solver can find an optimum when one exists and the model is feasible and bounded. GRG Nonlinear may find only a local optimum, and Evolutionary uses a heuristic search rather than guaranteeing a global answer.

Why is Solver missing from Excel?
The Solver Add-in may not be enabled. In Excel for Microsoft 365 on Windows, enable it through File > Options > Add-ins, choose Excel Add-ins in the Manage box, click Go, and select Solver Add-in. Excel for the web does not run this desktop Solver add-in.

Can Solver handle integer or binary variables?
Yes. You can add integer or binary constraints, but these restrictions can make a model much harder to solve. Larger or more complex discrete models may exceed the practical capabilities of the standard add-in even if they have fewer than 200 changing cells.

Why does Solver give different answers when I run it again?
Different starting values, nonlinear local optima, heuristic searches, or changes in workbook inputs can lead to different results. Check the selected method, Solver status, formulas, and constraints; for a nonlinear model, compare results from multiple starting points without assuming that any one run proves the global best.

Related

Try it: the Date Difference Calculator: days, months, years and weekdays between two dates, with the DATEDIF formulas.

More in How-to basics

Sources: Function behaviour and menu paths: Microsoft Support documentation for Excel. Method & sources

Spotted an error in this guide? Let us know.