diff --git a/src/fewmean.rs b/src/fewmean.rs
new file mode 100644
index 0000000..f634ccc
--- /dev/null
+++ b/src/fewmean.rs
@@ -0,0 +1,71 @@
+use num::{Float, FromPrimitive};
+use std::ops::{AddAssign, SubAssign};
+use crate::stats::Univariate;
+
+/// **Fading Exponentially Weighted Mean (FEWMean)**
+///
+/// This structure computes an **incrementally updated exponentially weighted mean** with a fading factor.
+///
+/// ## Difference from `ewmean.rs`:
+/// - `EWMean` is a classic **exponential moving average (EMA)** that applies a fixed smoothing factor (`alpha`).
+/// - The first few updates in `FEWMean` match the exact mean of a finite window of past values, whereas `EWMean` converges towards it **asymptotically**.
+/// - This is achieved by keeping track of a **decaying weight sum**, ensuring a more precise adaptation in early updates.
+///
+/// # Arguments
+/// * `fading_factor` - Controls the decay of past values. Smaller values give more weight to past data.
+///
+/// # Examples
+/// ```
+/// use watermill::fewmean::FEWMean;
+/// use watermill::stats::Univariate;
+///
+/// let mut few_mean: FEWMean<f64> = FEWMean::default();
+/// let data = vec![1., 3., 5., 4., 6., 8., 7., 9., 11.];
+/// for &x in data.iter() {
+///     few_mean.update(x);
+/// }
+/// assert_eq!(few_mean.get(), 9.4296875);
+/// ```
+#[derive(Clone, Copy, Debug)]
+pub struct FEWMean<F: Float + FromPrimitive + AddAssign + SubAssign> {
+    mean: F,
+    fading_factor: F,
+    weight_sum: F,
+}
+
+impl<F: Float + FromPrimitive + AddAssign + SubAssign> FEWMean<F> {
+    /// Creates a new `FEWMean` instance with a given fading factor.
+    pub fn new(fading_factor: F) -> Self {
+        Self {
+            mean: F::from_f64(0.0).unwrap(),
+            fading_factor,
+            weight_sum: F::from_f64(0.0).unwrap(),
+        }
+    }
+}
+
+impl<F> Default for FEWMean<F>
+where
+    F: Float + FromPrimitive + AddAssign + SubAssign,
+{
+    fn default() -> Self {
+        Self::new(F::from_f64(0.01).unwrap())
+    }
+}
+
+impl<F: Float + FromPrimitive + AddAssign + SubAssign> Univariate<F> for FEWMean<F> {
+    fn update(&mut self, x: F) {
+        if self.weight_sum == F::from_f64(0.0).unwrap() {
+            self.mean = x;
+            self.weight_sum = F::from_f64(1.0).unwrap();
+        } else {
+            let weight = (F::from_f64(1.0).unwrap() - self.fading_factor) * self.weight_sum;
+            self.mean = (weight * self.mean + x) / (weight + F::from_f64(1.0).unwrap());
+            self.weight_sum = weight + F::from_f64(1.0).unwrap();
+        }
+    }
+
+    fn get(&self) -> F {
+        self.mean
+    }
+}