For Admins
Call Spikes

Call Spikes

The portal continuously watches each company's call volume against its historical pattern. When a company receives unusually more calls than expected, subscribed admins are notified.

Reviewing call spikes

Open Call Spikes (opens in a new tab) from the Dashboard widget link. The page calculates spike windows from recording history and lets you filter by:

  • Company name
  • Company
  • Date range
  • Only anomalies

Each result shows the actual volume, expected volume, upper threshold, lower threshold, and z-score for that time window. Use View calls to open the affected company's Calls view for the same date.

The Dashboard call-spike widget also shows the current detection window, actual volume, expected volume, and upper threshold.

Investigating

A typical investigation flow:

  1. Open the Call Spikes page or the notification link.
  2. Look at the calls for the spike window — are they legitimate (real customers) or noise (one persistent robocaller, accidental flood from a marketing test, etc.)?
  3. Check whether the affected company recently launched a new ad campaign or service.
  4. Cross-reference with operational events — was Five9 having issues at that time?
  5. If real → notify the supervisor responsible so they can adjust staffing.
  6. If noise → consider blocking the source (handled outside the portal in Five9) and document the cause.

How forecasts and spikes relate

The forecasting feature predicts expected volume; the spike detection compares actual to expected. When you see a flagged spike, the forecast view shows you the gap visually.

If a customer's actual volume regularly diverges from forecast in a consistent direction, the model may need re-tuning for that customer's pattern. Note such cases for engineering review.

Spike severity rubric

A loose guide to triage urgency:

DeviationLikely meaningUrgency
Spike ≥3× normal during business hoursReal campaign or PR momentHigh — confirm staffing
Spike outside business hoursSpam wave or robocallerMedium — investigate, may need to block
Single-window deviationNoise or one-off eventLow — usually self-resolves

Coordinating with supervisors

Most spike alerts are best handled by the supervisor who owns the affected customer relationship. Your role as admin:

  • Investigate technical causes (Five9, S3, sync)
  • Tune the model when it consistently misfires
  • Escalate to engineering when it's not a customer-side issue

Next