Rolling average of daily active users

Question

Daily active users are noisy, so the product team reports a smoothed trend line instead.

Write a function `moving_average(values, window)` that returns a **list** of rolling averages, each rounded to 2 decimals. Use only full windows — a list of length `n` with window `k` returns `n - k + 1` values

Accepted Answer

Step 1: Convert to a float array and validate the window — a window of 0 or one longer than the data has no valid position. Step 2: `np.ones(window) / window` builds an averaging kernel: `window` weights that each equal `1/window`. Step 3: `np.convolve(arr, kernel, mode="valid")` slides that kernel across the array and keeps only positions where it fully overlaps — exactly the `n - k + 1` full windows requested. Step 4: Round to 2 decimals and call `.tolist()` to hand back plain Python floats. Step 5: In a pandas pipeline the equivalent is `series.rolling(window).mean().dropna()` — worth mentioning in an interview.

Rolling average of daily active users

More NumPy questions

Asked at Google