PlanMaker MMULT: Matrix Products, Output Shapes and Cost Models

The PlanMaker 2026 reference, 2024 reference and 2021 reference document this function. Original introduction and last-working-version endpoints remain unverified; a manual family is not a test of every build or edition.

References were retrieved on October 15, 2026. No native PlanMaker execution was performed. Examples below are independently calculated expectations, not application screenshots or native results. FreeOffice/mobile edition coverage and file transfers are not certified. English formulas use commas; verify separators and names in your installation.

=MMULT(Array1, Array2)

Both inputs are numeric cell ranges or arrays. The number of columns in Array1 must equal the number of rows in Array2. If the shapes are r×k and k×c, the result is r×c. The inner dimensions must match; the outer dimensions determine the output size.

Each result element is a row-by-column dot product. This differs from multiplying cells at matching positions. Matrix multiplication generally depends on order, and reversing the inputs can change both the dimensions and the values.

Create ProductControls with the following first matrix in A2:C3:

Put the second matrix in E2:F4:

The shapes are 2×3 and 3×2, so select H2:I3, a 2×2 output range. Enter =MMULT(A2:C3,E2:F4) and finish with Ctrl+Shift+Enter, as directed by the array-formula instructions. Expected results are:

For H2, calculate 1×7+2×9+3×11=58. For I3, calculate 4×8+5×10+6×12=154. The two shared dimensions count three products in every dot product; the output contains four cells.

The 2026 MMULT reference still requires an array formula. Select the full expected output before entering it; MUNIT’s separate dynamic Enter behavior is not a general instruction for MMULT. The manual warns that too small an output selection omits parts of the result, while surplus output cells can show #N/A.

For edits, select the complete existing array and finish again with Ctrl+Shift+Enter. The manual also describes Insert array formula on Android/iOS, but this guide has not independently verified mobile builds or performed native entry there.

In a copy change A2 from 1 to 2. Expected output rows become 65,72 and 139,154. Only the first output row changes, because the edit belongs to the first row of Array1. The increases 7 and 8 equal the first row of Array2.

The official example is =MMULT({1,2;3,4},{5,6;7,8}), giving rows 19,22 and 43,50. This is a separate 2×2 example, not the same matrices as ProductControls. Semicolons delimit rows in these US-style array constants; column separators depend on locale.

Using the original rectangular inputs, =MMULT(E2:F4,A2:C3) is also conformable, but its result is 3×3. Select a separate 3×3 output block and use array entry. Expected rows are 39,54,69; 49,68,87; 59,82,105.

The original product was 2×2. This example makes noncommutativity visible without relying only on an abstract statement that AB may differ from BA. For other matrix shapes, the reverse product might not even be defined.

On CostControls, enter two product rows 2,1,3 and 1,4,2 in A2:C3. Put the matching unit prices 5,10,2 in E2:E4, one price for each input column. Select G2:G3 and enter =MMULT(A2:C3,E2:E4) as an array formula. Expected totals are 26 and 49.

For a second price scenario, add 6,9,3 in F2:F4. Select I2:J3 and use =MMULT(A2:C3,E2:F4). The expected result has rows 26,30 and 49,48. Each output column is one complete pricing scenario; each output row belongs to one product row.

The mathematics cannot detect a price list in the wrong business order. Keep item labels outside the numeric ranges and verify that each price row corresponds to the same item as its quantity column.

Text, headers and nonconforming shapes violate the documented numeric matrix contract; exact native error tokens for these inputs are not supplied by the short reference. Blank coercion, maximum dimensions and arithmetic overflow are also unverified.

Check shapes before changing cell values, then verify one dot product manually. If a result seems transposed, trace the row/column roles rather than swapping ranges at random. Record the build and locale, and test saved-file behavior separately from these independent arithmetic expectations.