# Conversation-Statistics Workbook — Advanced Statistical Analytics The Conversation-Statistics workbook is the fifth and final workbook in the Conversations gallery. It goes beyond volume and disposition reporting to provide advanced statistical and telecommunications-grade metrics for conversation performance analysis. Where other workbooks answer “how many” and “what type,” this workbook answers “how variable,” “how extreme,” and “how much capacity do we need?” The workbook contains eight sub-views: Dashboard | Moving Average of Conversations | Shortest Conversation (secs) | Longest Conversation (secs) | Standard Deviation of Conversations | BHCA | Erlang | Details Page. The Dashboard sub-view (Image 1) presents all six statistical indicators simultaneously as stacked time-series panels, providing a comprehensive snapshot of platform behavior across the full reporting window.

Figure 1: Conversation-Statistics Dashboard — Six Statistical Indicators (Tue 3/10/26 – Tue 4/7/26, Employee Platform)

### 1. Metric 1 — Moving Average of Conversations The Moving Average of Conversations shows a smoothed trend line of daily conversation volume, calculated as a rolling average across the selected period. Unlike the raw daily count (which shows day-to-day fluctuations), the moving average smooths out noise to reveal the underlying growth or decline trend in platform usage. Data visible in the dashboard: The moving average begins at 5.75 (Fri 3/13/26) and remains flat and low through mid-March. Starting from Mon 3/30/26 it climbs steeply, reaching 7,950.38 by Tue 3/31/26 — driven by the massive conversation spike documented in Section 26.3 (53,595 conversations on 3/31). The line then plateaus at a high level through early April. The moving average reaching 7,950 is not a reflection of sustained daily volume — it is a mathematical artifact of the 53,595-conversation spike on 3/31 being incorporated into the rolling window. This illustrates an important limitation of moving averages: a single extreme event can distort the trend line for the duration of the lookback window. In production environments with stable volumes, the moving average provides a reliable trend signal. Available filters: Date Selection Type, Start Date, End Date, Channel, Stage, Assistant, Business Day. Business value: Moving average trending is the standard tool for presenting conversation volume growth to leadership. It eliminates the visual noise of weekday/weekend cycles and enables clear identification of whether platform usage is growing, stable, or declining — independent of individual day anomalies. ### 2. Metric 2 — Shortest Conversation (secs) This panel tracks the duration of the shortest (minimum) conversation recorded each day. It directly corresponds to the “Min” component of the Max, Avg and Min Durations view in Section 27.1.5. Data visible in the dashboard: Multiple dates show 0 secs as the shortest conversation — specifically Tue 3/24/26, Wed 3/25/26, Mon 3/30/26, and Tue 3/31/26. A 0-second conversation means a session was created and immediately terminated, with no measurable interaction time recorded. A persistent floor of 0 seconds is an important quality signal. It indicates that on those dates, at least one conversation in the dataset had essentially no duration — the session opened and closed instantaneously. This can result from: authentication failures that immediately terminate the session; bot initialization errors that crash before the first interaction; automated health check pings that create and immediately close sessions; or data recording issues where the start and end timestamps are identical. Investigating the specific ConversationIds with 0-second durations (via the Details Page) is the recommended remediation step. Available filters: Date Selection Type, Start Date, End Date, Channel, Stage, Assistant, Business Day. ### 3. Metric 3 — Longest Conversation (secs) This panel tracks the duration of the longest (maximum) conversation recorded each day — the single conversation that took the most time on a given date. Data visible in the dashboard: The peak is 2,698 seconds (approximately 45 minutes) occurring around Tue 3/24/26. A secondary peak of approximately 1,800 seconds (30 minutes) appears around Fri 3/27/26. Both values decline through early April, settling at lower levels by 4/2/26. A longest conversation of 2,698 seconds (44.9 minutes) in an AI assistant context is extremely high and almost certainly represents a conversation where: the user left the session open without interacting (session remained active until timeout); the assistant entered a loop state where it repeatedly prompted without resolution; or the conversation involved an unusually complex multi-step flow with extended pauses. Cross-referencing this date and duration with the Conversation Viewer (using the Details Page to find the specific ConversationId) would reveal the exact interaction that drove this maximum. Available filters: Date Selection Type, Start Date, End Date, Channel, Stage, Assistant, Business Day. Business value: Monitoring the daily maximum duration helps set and enforce conversation timeout policies. If the maximum regularly exceeds a defined threshold (e.g., 10 minutes for a chat assistant), it indicates the timeout configuration may be too permissive, or specific conversation flows have runaway branches that need correction. ### 4. Metric 4 — Standard Deviation of Conversations Standard Deviation measures the variability in daily conversation counts — how much individual day volumes deviate from the overall average. A low standard deviation indicates consistent, predictable daily volumes; a high standard deviation indicates highly variable traffic that is difficult to predict and plan for. Data visible in the dashboard: The standard deviation spikes sharply to 6,121.0 around Fri 3/27/26, then drops back. The dataset average is shown as Avg = 489.5. The value of 0.0 appears at the end of the period (early April), indicating that on those days the daily conversation count was near-identical to the period average. The spike to 6,121 on 3/27 reflects the extreme volume variability introduced by the 53,595-conversation day on 3/31 (documented in Section 26.3). When that outlier day is included in the standard deviation calculation, it creates an enormous variance signal. This is mathematically expected: one extreme outlier dramatically inflates standard deviation. In production without such outliers, standard deviation provides a reliable measure of traffic predictability. Available filters: Date Selection Type, Start Date, End Date, Channel, Stage, Assistant, Business Day. Business value: Standard deviation is the primary metric for infrastructure capacity planning under uncertainty. Teams use it to calculate confidence intervals for traffic projections: if average daily volume is 5,000 conversations with a standard deviation of 500, then provisioning for 6,500 (average + 3 standard deviations) covers 99.7% of expected days. High standard deviation requires over-provisioning or auto-scaling to handle peak variability. ### 5. Metric 5 — BHCA (Busy Hour Call Attempts) BHCA (Busy Hour Call Attempts) is a telecommunications-grade metric that identifies the single hour within each day that experiences the highest volume of conversation attempts. It is expressed as the count of attempts during that peak hour, with the specific hour labeled in parentheses (e.g., “19,500 (5 AM)” means 19,500 attempts occurred during the 5 AM hour). Data visible in the dashboard: The peak BHCA value is 19,500 (5 AM) on Tue 3/31/26 — consistent with the overall volume spike on that date. Earlier dates show minimal BHCA values of 2 (3 PM on Thu 3/12/26), 2 (8 AM on Mon 4/6/26), and 2 (12 PM on Mon 4/6/26). The 5 AM peak on 3/31 confirms the spike was concentrated in early morning hours, strongly suggesting automated/programmatic activity rather than organic user behavior (human users rarely generate 19,500 conversations at 5 AM). Available filters: Date Selection Type, Start Date, End Date, Channel, Stage, Assistant, Business Day. Business value: BHCA is the standard metric for telecommunications capacity engineering. It drives infrastructure sizing decisions: the platform must be provisioned to handle at least the BHCA volume per hour without degradation. Knowing both the BHCA value AND the hour at which it occurs enables targeted scaling strategies — for example, scheduling auto-scaling to activate before 5 AM if that is consistently the peak hour, or adjusting rate limiting policies to protect system stability during peak windows. ### 6. Metric 6 — Erlang Erlang is a dimensionless telecommunications unit that measures traffic intensity — specifically, the average number of concurrent conversations in progress at any given moment during the measurement period. It is calculated as: Erlang = (Number of calls per hour × Average call duration in hours). An Erlang value of 1.0 means that on average, exactly one conversation is always in progress. A value of 5.3 means an average of 5.3 simultaneous conversations are ongoing at any moment. Data visible in the dashboard: The Erlang value peaks at 5.3 on Mon 3/30/26, then drops sharply to 0.0 by Wed 4/1/26. This pattern mirrors the BHCA and volume spikes: as conversation volume surged on 3/30–3/31, concurrent load increased to 5.3 simultaneous conversations on average. The return to 0.0 confirms that by 4/1, the spike had fully subsided. The dedicated Erlang by Date sub-view (Image 2) shows this metric in isolation with full filter controls. The filter panel for this sub-view confirms: Channel = Chat (important — Erlang is calculated for the chat channel specifically in this view), Stage = STG, Assistant = (None), Business Day = (All). Available filters: Date Selection Type, Start Date, End Date, Channel, Stage, Assistant, Business Day. Business value: Erlang is the foundational metric for calculating how many concurrent processing threads, server instances, or agent seats a platform needs. In traditional call center operations, Erlang B and Erlang C formulas are used to calculate required agent staffing from traffic intensity. For AI assistant platforms, Erlang translates directly into concurrent session capacity requirements: if peak Erlang is 5.3, the platform needs capacity for at least 6 simultaneous sessions at any moment. In production at scale, Erlang values in the hundreds or thousands drive cloud infrastructure provisioning decisions. ### 7. Sub-View Reference — Dedicated Views
Sub-ViewDescription and Available Filters
Moving Average of ConversationsDedicated full-page view of the rolling average trend line. Filters: Date Selection Type, Start/End Date, Channel, Stage, Assistant, Business Day.
Shortest Conversation (secs)Dedicated full-page view of daily minimum duration. Filters: Date Selection Type, Start/End Date, Channel, Stage, Assistant, Business Day. Use to detect 0-second session anomalies.
Longest Conversation (secs)Dedicated full-page view of daily maximum duration. Filters: Date Selection Type, Start/End Date, Channel, Stage, Assistant, Business Day. Use to enforce timeout policies.
Standard Deviation of ConversationsDedicated full-page view of daily volume variability. Filters: Date Selection Type, Start/End Date, Channel, Stage, Assistant, Business Day. Use for capacity planning confidence intervals.
BHCADedicated full-page view of busy hour call attempts with peak hour labeling. Filters: Date Selection Type, Start/End Date, Channel, Stage, Assistant, Business Day. Use for hour-level infrastructure sizing.
ErlangDedicated full-page view of traffic intensity (concurrent sessions). Filters: Date Selection Type, Start/End Date, Channel (Chat by default), Stage, Assistant, Business Day. Core metric for concurrency capacity planning.
Details PageRecord-level drill-down providing individual conversation records with full metadata. Used to investigate extreme values (e.g., the specific conversation that generated 2,698 seconds maximum duration on 3/24/26).
### 8. Business Value — Why This Workbook Matters The Conversation-Statistics workbook is the analytical foundation for infrastructure capacity planning, SLA compliance, and operational risk management. While other workbooks describe what is happening conversationally (volume, disposition, handle time), this workbook describes the statistical properties of that activity — how predictable it is, how extreme the outliers are, and how much concurrent capacity is required. The six metrics work together as a system: the Moving Average establishes the baseline trend; Standard Deviation quantifies the uncertainty around that trend; Shortest and Longest Conversations bound the duration distribution; BHCA identifies the peak demand the infrastructure must withstand in any single hour; and Erlang translates that peak demand into a concrete concurrent capacity requirement. Together, these metrics enable data-driven infrastructure decisions rather than rule-of-thumb provisioning. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.ixhello.com/ix-hello-reporting/premium-reporting/total-conversations-with-contained-and-transferred-dedicated-sub-view/conversation-statistics-workbook-advanced-statistical-analytics.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.