LOGEST Function in Microsoft Excel: Syntax, Examples & Troubleshooting
Estimate Excel exponential-model coefficients with LOGEST. Interpret multiplier versus growth rate, reversed predictor order, const and the five-row statistics output.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
Quick answer: LOGEST fits an exponential curve and returns the coefficients describing it. With one predictor, the coefficient row is m, b for y = b×m^x. A returned m of 2 means doubling per x unit, equivalent to a 100% increase, not a 2% increase.
Syntax and defaults
=LOGEST(known_y's, [known_x's], [const], [stats])
Known y values are required. Known x is optional and defaults to 1,2,3,…. Use positive y values for the exponential/logarithmic fit. const TRUE or omitted estimates b; FALSE fixes b = 1 and refits the multipliers. stats FALSE or omitted returns coefficients only; TRUE returns additional statistics.
With multiple predictors, the model is b×m1^x1×m2^x2×…, and the returned coefficient row is mn,…,m1,b. The order reverses the predictor columns; it is not the input column order.
Worked example: recover a known exponential curve
Enter x values 0,1,2,3 in A2:A5 and corresponding y values 3,6,12,24 in B2:B5.
- Enter
=LOGEST(B2:B5,A2:A5,TRUE,FALSE)in D2, with D2:E2 available. - Expected coefficient row: 2,3, within normal floating-point precision.
- The reconstructed model is y = 3×2^x.
- At x = 4, independent substitution gives 48.
The first result is the multiplier, and the second is the value at x=0. Reversing them would yield a different curve. At x=2, the correct reconstruction gives 3×2² = 12, matching the input.
Second example: a unit-baseline fit
Put x values 1,2,3 in H2:H4 and y values 2,4,8 in I2:I4. =LOGEST(I2:I4,H2:H4,FALSE,FALSE) has expected coefficient row 2,1. The FALSE constraint is consistent with y=2^x.
It is not consistent with the first dataset’s baseline of 3. A const FALSE result is a refitted model under a different constraint, not just the earlier coefficient row with its final entry edited.
Advanced application: read the statistics layout correctly
With stats TRUE, the output has five rows. For one predictor, reserve a 5-by-2 area:
| Row | First column | Second column |
|---|---|---|
| 1 | m coefficient | b constant |
| 2 | Coefficient standard error | Constant standard error |
| 3 | R² | Estimate standard error |
| 4 | F statistic | Degrees of freedom |
| 5 | Regression sum of squares | Residual sum of squares |
With more predictors, the first two rows contain additional coefficient columns in reverse order; the remaining diagnostic pairs occupy the left side of the array. Do not mistake padding/unused entries for additional model coefficients.
An exponential relationship becomes linear in logarithms: ln(y) = ln(b)+x×ln(m). This is why an exponential fit and an ordinary raw-y straight-line fit answer different questions. Interpret diagnostics in the context of that transformed model, rather than treating them automatically as raw-unit prediction intervals. An exact synthetic exponential dataset has no real residual variation; this guide does not promise a finite F statistic or native error value for every perfect-fit diagnostic.
For a practical model, compare reconstructed predictions with observations and inspect residual patterns. The coefficients alone do not show that a forecast will hold outside the observed range. Microsoft explicitly cautions about extrapolated predictions. A high fit statistic does not establish causation or remove uncertainty.
Array entry and data-shape checks
In a current Microsoft 365 build with dynamic arrays, enter the formula in the top-left output cell and press Enter, leaving enough empty cells for its result. In a legacy array workflow, select the intended output range first and confirm the formula with Ctrl+Shift+Enter. Excel inserts the braces; do not type them manually.
Keep each observed y paired with the corresponding x row. A single predictor requires matching x/y dimensions. For multiple predictors, a column of y observations uses one x column per predictor; a row of y observations uses one x row per predictor. New prediction inputs must retain the same predictor layout. Headers are not observations.
A blocked dynamic-array destination can produce #SPILL!, while incomplete legacy output selection can hide results. The reviewed function page does not provide a complete invalid-shape/error/coercion table, so no cross-application error behavior is assumed here. Inspect missing, text and erroneous source values rather than silently replacing them with zeros.
Compatibility and verification
Microsoft’s reference checked on October 17, 2026 lists: Excel for Microsoft 365 Excel for Microsoft 365 for Mac Excel 2024 Excel 2024 for Mac Excel 2021 Excel 2021 for Mac Excel 2019 Excel 2016. This current support list does not establish the first release. The introduction version and last functioning version have not been established from the reviewed documentation. No retirement is asserted.
Example verification: Expected numerical results are independently calculated; text and shape expectations follow the documented contract. These examples have not been tested in Microsoft Excel. No screenshot or native execution is claimed.
Related guides and source
LN in Excel | EXP in Excel | POWER in Excel | INDEX in Excel
- Microsoft LOGEST reference, retrieved October 17, 2026.