cx-contact-spike-detectionlisted

# Detecting contact spikes Support volume is one of the more strongly seasonal series in a business, and almost every naive spike detector fails on the same thing: it compares a Monday to a Sunday and declares an incident. Done properly, support contacts are among the fastest incident signals a company has — customers frequently notice a failure before monitoring does. That is the prize, and it is only reachable with a baseline that models the periodicity. ## Model the seasonality, or the detector is noise At minimum, support traffic has: - **A day-of-week cycle**, typically large. Weekday and weekend volumes can differ by more than any incident you are trying to catch. - **An intraday cycle**, larger still, and it differs by channel and by market. - **Public holidays**, which differ per market and break every baseline. A holiday calendar is not optional, and a multi-market operation needs one per market. - **Business-driven cycles** — pay days, billing dates, statement runs, campaign sends, release days. These are predictable and produce large recurring spikes that a naive detector will flag every month. - **Trend**, from growth in the customer base. **Compare like with like**: same weekday and same hour band, over several recent weeks, or use a baseline that explicitly models the period. Comparing to "the last 7 days average" bakes the weekly cycle into the residual and produces a detector that fires every Monday and every Saturday. ## Rate and count, togethe