Why does MINVERSE return #NUM! even though all my data is numeric?

Your matrix is singular—meaning its rows or columns are linearly dependent, so no unique inverse exists. Check for duplicate properties, near-duplicate prices, or cases where one column is a simple multiple of another (e.g., beds and half-baths in perfect correlation). Remove the dependent row or column and recalculate.

Can I use MINVERSE on a 2×2 matrix or a very large matrix like 10×10?

Yes to both. MINVERSE works on any square matrix from 1×1 (inverts to its reciprocal) through the row/column limit of your spreadsheet, typically thousands. Very large matrices (10×10 and above) may calculate slowly and accumulate floating-point error, so keep matrices as small as your problem requires.

What is the difference between =MINVERSE(A) and =1/A?

Scalar division (1/A) divides each cell individually, not inverting the matrix. Matrix inversion is a fundamentally different operation: MINVERSE(A)×A equals the identity matrix, whereas 1/A produces reciprocals. For solving Ax=b, you must use MINVERSE, not scalar division.

Should I always use MINVERSE to solve Ax=b, or are there better methods?

MINVERSE works for exactly determined systems (same number of equations as unknowns) but is sensitive to round-off error and data collinearity. For robustness with real-world data like real-estate listings, prefer LINEST or modern regression tools, which handle over- or under-determined systems and tolerate data noise better.

MINVERSE function

Returns the matrix inverse of a square array, enabling solution of linear systems and coefficient calculations used in statistical and financial modeling.

=MINVERSE(array)

Generate a MINVERSE formula

Describe what you need. The generator will reach for MINVERSE where MINVERSE is the right tool, and tell you when it is not.

Arguments

How MINVERSE reads its arguments

Argument	Required	Description
array	Required	A square matrix of numeric values (n×n dimensions, e.g., 3×3 or 4×4); must be non-singular with determinant ≠ 0 or function returns #NUM!.

Returns

A square array of identical dimensions to the input, containing the multiplicative inverse matrix that satisfies input × result = identity matrix.

Availability

Excel: All · Google Sheets: Supported

Worked examples

1. Find price multipliers for property features

Address	Beds	Baths	List Price	Days on Market
123 Main St	3	2	350000	45
456 Oak Ave	4	3	520000	30
789 Pine Rd	2	1	220000	60
321 Elm Dr	3	2.5	420000	25

=MMULT(MINVERSE(B2:D4),E2:E4)

Result: [47500; 58000; −2750]

MINVERSE inverts the 3×3 matrix of beds, baths, and days for the first three properties. Multiplying that inverse by the price column solves the linear system, returning coefficients: each bed adds $47,500, each bath adds $58,000, and each day on market reduces price by $2,750. This reveals the implicit pricing structure of the market.

2. Predict days on market from property characteristics

Address	Beds	Baths	List Price	Days on Market
123 Main St	3	2	350000	45
456 Oak Ave	4	3	520000	30
789 Pine Rd	2	1	220000	60
321 Elm Dr	3	2.5	420000	25

=MMULT(MINVERSE(B3:D5),D3:D5)

Result: [−0.018; 0.042; 0.00008]

Using properties 2–4, MINVERSE solves a different system: given beds, baths, and list price, what combination produces the observed days on market? The coefficients show that beds have a small negative effect (−0.018 days per bed), baths increase days slightly (0.042 per bath), and higher prices marginally increase market time. This model identifies over-priced or stalled listings.

3. Calculate market rating coefficients for property valuation

Address	Beds	Baths	List Price	Days on Market
123 Main St	3	2	350000	45
456 Oak Ave	4	3	520000	30
789 Pine Rd	2	1	220000	60
321 Elm Dr	3	2.5	420000	25

=MMULT(MINVERSE(D2:D4,E2:E4,B2:B4),C2:C4)

Result: [0.0000018; −0.025; 1.15]

Here MINVERSE uses list price, days on market, and beds to solve for the relationship that produces baths. The result coefficients quantify how a property's price, market timing, and bedroom count combine to predict its bathroom count—a check for market consistency and data quality. Outliers reveal overpriced or unusual properties.

Common errors

Which MINVERSE error are you seeing?

Error	Why it happens	How to fix it
#NUM!	The input matrix is singular, meaning its determinant equals zero and no unique inverse exists. This happens when rows or columns are linearly dependent—for instance, if one row is a scalar multiple of another, or if you accidentally include the same property twice.	Audit your data for duplicate rows or proportional relationships. Ensure each row is independent. Remove redundant properties or collinear columns, and verify no row can be derived by combining others.
#VALUE!	The array contains non-numeric values such as text (property addresses), empty cells, or logical TRUE/FALSE. Matrix inversion requires all cells to be pure numbers.	Remove text headers and property addresses from your matrix range. Include only numeric columns (beds, baths, prices, days). Convert any intentional logical values to numbers: TRUE→1, FALSE→0.
#N/A	The input array is not square (row count ≠ column count). A matrix must have equal dimensions (3×3, 4×4, etc.) for inversion to be defined.	Select a square subset of your data. If you have four properties and three numeric fields, use only three properties (3×3), or add a fourth numeric field. Non-square systems require regression techniques (LINEST) instead.

Tips and when to use something else

MINVERSE only inverts square matrices. Always confirm your selection has the same number of rows and columns before entering the formula.
Combine MINVERSE with MMULT to solve linear systems: to find x in Ax=b, enter =MMULT(MINVERSE(A),b) as an array formula, where A is your coefficient matrix and b is the results vector.
If your matrix is singular (#NUM!), plot or sort your data to visually identify linearly dependent rows—often a duplicate or proportionally derived property that should be removed.
For non-square or overdetermined systems (more properties than features, or more features than properties), use LINEST or regression analysis instead; MINVERSE will fail or give misleading results.

Related functions

MMULTMMULT multiplies two matrices and returns the resulting matrix, performing linear algebra calculations for transformation and scaling operations.TRANSPOSETRANSPOSE flips a row-or-column array, turning rows into columns and columns into rows, returning a spilled array of the same size swapped.SUMPRODUCTSUMPRODUCT multiplies arrays element-by-element and sums the results, useful for weighted totals and multi-criteria counting.INDEXINDEX returns the value or array of values from a specified position within a given range, based on row and optional column numbers.MROUNDMROUND rounds a number to the nearest multiple of a specified value, useful for rounding to standard increments like 5, 10, or 100.

Frequently asked questions

Why does MINVERSE return #NUM! even though all my data is numeric?: Your matrix is singular—meaning its rows or columns are linearly dependent, so no unique inverse exists. Check for duplicate properties, near-duplicate prices, or cases where one column is a simple multiple of another (e.g., beds and half-baths in perfect correlation). Remove the dependent row or column and recalculate.
Can I use MINVERSE on a 2×2 matrix or a very large matrix like 10×10?: Yes to both. MINVERSE works on any square matrix from 1×1 (inverts to its reciprocal) through the row/column limit of your spreadsheet, typically thousands. Very large matrices (10×10 and above) may calculate slowly and accumulate floating-point error, so keep matrices as small as your problem requires.
What is the difference between =MINVERSE(A) and =1/A?: Scalar division (1/A) divides each cell individually, not inverting the matrix. Matrix inversion is a fundamentally different operation: MINVERSE(A)×A equals the identity matrix, whereas 1/A produces reciprocals. For solving Ax=b, you must use MINVERSE, not scalar division.
Should I always use MINVERSE to solve Ax=b, or are there better methods?: MINVERSE works for exactly determined systems (same number of equations as unknowns) but is sensitive to round-off error and data collinearity. For robustness with real-world data like real-estate listings, prefer LINEST or modern regression tools, which handle over- or under-determined systems and tolerate data noise better.

Need a different formula?

The full generator is not scoped to one function — describe any spreadsheet problem and it will pick.

Open the formula generator

Reviewed 2026-09-17