Compliance Alert / Trailing-Average Sliding Window

A compliance system raises an alert whenever the trailing-window average of call volumes exceeds a threshold. Count how many such alerts are raised over the full input. A sliding-window staple that recurs in Goldman's OA bank. Requirements A compliance system monitors incoming and outbound calls. At each time T, average the call volumes over the trailing precedingMinutes minutes, i.e. T - (precedingMinutes - 1) through T. An alert is sent if this average is strictly greater than alertThreshold. No alerts are sent until at least T = precedingMinutes because there are not enough values to consider yet. Return the total number of alerts raised over the full input. public static int calculateAlert(int[] numCalls, int alertThreshold, int precedingTime) Examples n = 8 numCalls = [2, 2, 2, 2, 5, 5, 5, 8] alertThreshold = 4 precedingMinutes = 3 No alerts until T = 3. From T = 3 to T = 8, the trailing averages are 2, 2, 3, 4, 5, 6. Values strictly greater than 4: 5 and 6 → 2 alerts. Notes Single sliding-window pass: maintain a running sum; on each step add numCalls[i] and subtract numCalls[i - precedingTime] once the window is full. Off-by-one is the main trap: the first valid window ends at index precedingTime - 1, not at precedingTime. Make sure the very first window is also evaluated against the threshold. Use integer comparison sum > alertThreshold * precedingTime rather than floating-point division to avoid precision issues. Goldman's hidden tests do not always reveal this, but the safer form is what interviewers expect when asked to discuss complexity / precision. O(n) time, O(1) space. Preparation Write the brute-force O(n * precedingTime) form first, then refactor to the rolling sum — interviewers like to see both. Replace the float comparison with the integer-product form and walk through why floating point would matter at scale. Practice closely-related LC: "Maximum Average Subarray I" (LC 643) is the same skeleton with a different return.