Excel Solver Tolerance Settings Explained

How-to basics · Updated November 21, 2026

In Excel Solver for Microsoft 365 on Windows, open Data > Solver, choose Solver Parameters > Options, and distinguish Precision from Tolerance: Precision controls how closely constraints must be met, while Tolerance is used with integer-constrained problems to set how close Solver’s best integer result must be to its optimality bound before it can stop. Tightening either setting can take longer, so adjust the one that matches the issue rather than lowering every number.

Where to find the settings

On Windows, open the Data tab and select Solver. In the Solver Parameters dialog, select Options to see the solver’s calculation settings. If Solver is missing from the Data tab, enable the add-in through File > Options > Add-ins. At the bottom of the window, choose Excel Add-ins from the Manage list, select Go, check Solver Add-in, and select OK. Then return to Data > Solver.

On a Mac, enable the add-in from Tools > Excel Add-ins, select Solver Add-in, and open it from the Data tab. Menu wording and available options can differ slightly by Excel build. Excel for the web does not provide the same built-in Solver workflow as the desktop app; open the workbook in desktop Excel to configure and run this Solver add-in.

The Options dialog contains more than tolerance controls. It also includes limits such as maximum time and iterations, plus settings that affect how Solver estimates and searches for a solution. Those settings address different aspects of the calculation; changing them is not a substitute for choosing a suitable tolerance.

Precision and integer optimality tolerance are different

Precision concerns whether a solution satisfies the model’s constraints closely enough. For example, if a production constraint says total material use must be no more than 100 units, precision affects how closely Solver must respect that boundary in its numerical calculations. A smaller precision value asks for tighter constraint handling. It does not mean that Solver will find a better objective value, and it does not turn a poorly specified model into a sound one.

Tolerance, often described as integer optimality tolerance, applies to problems with integer or binary decision variables. Solver may find a feasible integer solution and compare its objective value with a bound on the best value that could still be reached. The tolerance is the permitted percentage gap between those values. A smaller percentage generally asks Solver to keep searching longer for a result closer to the bound; it does not permit a variable declared integer to be fractional.

These settings therefore answer separate questions: “Does this result satisfy the constraints closely enough?” and “How close to proven optimality must the integer result be before the search can stop?” If a reported solution violates a business rule, inspect the constraints and Precision. If the result is feasible but you want a stronger guarantee about its objective value, review integer Tolerance and the Solver report.

What tolerance means in an integer example

Imagine a delivery plan that chooses whole numbers of vans and seeks to minimize total cost while meeting demand. The van-count variables must be integers, so Solver cannot return 3.4 vans as a valid final decision. During its search, however, it can use bounds to estimate how good a still-unfound solution might be. The integer optimality tolerance determines how large a gap between the best feasible integer plan and the relevant bound Solver may accept when stopping.

If the tolerance is relatively loose, Solver may stop with a valid plan sooner even though its search has not closed the gap as tightly. A tighter tolerance can produce a stronger optimality result, but may require more computation and still may not finish within the time or iteration limits. The tolerance is not a promise that the plan’s cost will improve by that percentage when you tighten it; it is a stopping criterion based on the search’s gap.

This distinction matters when interpreting results. A feasible plan is not automatically a proven global optimum, and a small tolerance does not guarantee that Solver can complete the proof. Check the Solver results message and reports, and record the objective value and decision variables before comparing runs. If a tighter run reaches the same plan, that is useful evidence, but the Solver status remains important.

How to choose a setting

Begin with the default values and a correctly constructed model. A sensible setting depends on the scale and purpose of the problem; there is no single tolerance that is best for every workbook. Consider these factors:

  • Consequences of a small constraint miss: If even a tiny violation would make a result unusable, review the relevant constraint, units, and Precision. Do not assume that an extremely small value is always practical; numerical calculations and the scale of the inputs matter.
  • Whether decisions must be whole numbers: Integer Tolerance is relevant when integer or binary restrictions are part of the model. If every decision variable is continuous, changing this integer-specific setting will not make the continuous solution more accurate.
  • Time available: A tighter integer optimality target can increase the search burden. Balance the value of additional certainty against the time limit and the complexity of the model.
  • Meaningful measurement units: A model expressed in cents, dollars, thousands of dollars, or millions can behave differently numerically. Keep units consistent and check whether very large and very small coefficients appear together.
  • Required confidence: A classroom exercise, planning estimate, and high-stakes operational decision may call for different verification. Use independent checks and appropriate review rather than treating a tolerance value as a quality certification.

For a small integer model, try a modestly tighter integer tolerance only if the result needs a closer optimality bound and Solver can still finish. For a model that is simply failing a constraint, investigate Precision and the constraint formulation instead. Change one setting at a time, rerun the model, and compare both the result and Solver’s completion message so you can tell what changed.

Precision is not the same as rounding

A common misconception is that Precision controls how many decimal places appear in a cell. It does not. Number formatting changes what Excel displays, not the underlying value used in the calculation. Likewise, using ROUND in a worksheet changes a model only if the rounded result is actually part of the formulas or constraints; displaying a number with zero decimal places does not make it an integer decision.

If a decision must be a whole number, add an integer constraint in Solver rather than relying on cell formatting. If a value may be continuous but should display neatly, format the cell separately and preserve the full calculated value for Solver. For example, a quantity shown as 10 may internally be 9.999999, and its display alone cannot establish whether the constraint is satisfied. Check the actual cell value and model requirements.

Also distinguish Solver’s numeric precision from the precision of the source data. If demand is estimated to the nearest hundred units, setting a very strict numerical threshold cannot make that estimate more accurate. The tolerance controls Solver’s calculation and stopping behavior; it cannot correct inaccurate inputs or assumptions.

Common problems and practical checks

If Solver reports a solution that appears to break a limit, first inspect the actual cell values and the exact constraint references. A constraint may point to the wrong cell, omit part of a total, or compare cells that use inconsistent units. Confirm that the worksheet recalculates as expected and that the changing cells, objective cell, and constraints all refer to the intended model.

If a tighter setting makes Solver run much longer, restore the previous value and consider whether the model itself can be improved. Remove unnecessary decision variables or redundant constraints, use reasonable bounds, and avoid formulas with abrupt or discontinuous behavior unless the selected solving method is appropriate. Increasing the maximum time or iteration limit may allow more searching, but it does not ensure that Solver will find a better solution.

For integer problems, check that the integer restrictions are attached to the correct changing cells and that the chosen solving method supports the model. The Simplex LP method is intended for linear models; GRG Nonlinear handles smooth nonlinear models; Evolutionary is designed for models with more difficult, discontinuous, or nonsmooth relationships. Method choice can affect whether Solver reaches a useful result, independently of the tolerance setting.

Keep a record of the original settings and the results before experimenting. Solver can overwrite changing-cell values when it finds a solution. Use Solver’s Restore Original Values option when appropriate, or preserve a copy of the workbook or starting values before a comparison run.

How to make a reliable adjustment

  1. Save a copy of the workbook and note the current objective value, changing-cell values, Solver method, and key settings.
  2. Run Solver once with the existing settings. Read the completion message and inspect the solution rather than judging it only by formatted cell displays.
  3. Identify the specific concern: constraint satisfaction, integer optimality, or an overall search that has run out of time or iterations.
  4. Change only the matching setting. For a constraint concern, consider Precision and verify the model. For an integer optimality concern, adjust Tolerance. For a time-limit message, review model complexity and limits as well.
  5. Run again and compare the objective, feasibility, and completion status. Keep a change only if it serves the model’s actual requirement.

This controlled process makes tolerance experiments interpretable. If several options change at once, a different result does not reveal which change mattered. For important decisions, independently recalculate key totals and have the model reviewed; a Solver tolerance is a computational setting, not a substitute for validation.

Frequently Asked Questions

What is the best tolerance setting for Excel Solver?
There is no universal best value. Start with the default, then tighten the integer optimality tolerance only when a closer optimality bound is worth additional search time. Use Precision when the concern is whether constraints are met, and verify that the model’s inputs and units are appropriate.

Does a lower Solver tolerance mean more accurate answers?
For an integer model, a lower integer optimality tolerance requests a smaller gap to the bound before Solver stops. It can increase search time and does not guarantee a different or better feasible plan. Lowering Precision instead concerns constraint satisfaction, not the objective gap.

Does Solver tolerance allow a constraint to be violated?
Integer optimality tolerance is not permission to ignore a constraint. Precision governs how closely constraints are handled numerically, while the model’s constraints still define what is feasible. Check the actual values, constraint references, and Solver status if a result appears infeasible.

Should I change tolerance for a linear Solver model?
Only if the model has integer or binary decision constraints and you need to change the integer optimality stopping criterion. A continuous linear model does not gain a better integer solution from that setting. For constraint concerns, review Precision and the formulation instead.

Why does Solver take longer after I lower tolerance?
A stricter integer optimality target can require Solver to explore more possibilities before closing the gap to its bound. Integer search may grow demanding as the model gains variables and constraints. Check the time and iteration limits, but also consider simplifying or validating the model rather than merely allowing it to run longer.

Does changing cell decimal places change Solver tolerance?
No. Cell formatting affects display only. Solver uses the underlying values and its own options. Use integer constraints when decisions must be whole numbers, and adjust the relevant Solver setting separately.

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.