CONTACT CENTRE PERFORMANCE - PUBLIC AGGREGATE DATA

This release accompanies a portfolio case study. It contains observed aggregate statistics, not individual interaction records or synthetic data.

FILES
public-aggregate-data.json: chart data, exact metric definitions and release policy.
summary.csv: headline aggregates, one row.
periods.csv: six reporting periods in chronological order; the last is partial.
channels.csv: recorded inbound, outbound and chat volumes.
known_outcomes.csv: the four major known inbound outcomes.
team_operations.csv: operations by generic initial routing team.
typology_families.csv: seven broad families by channel.

UNITS AND DEFINITIONS
Rates use decimals from 0 to 1, so 0.6839 means about 68.39%. NPS ranges from -100 to 100. Satisfaction is a mean score out of 10.
Known human-routed calls = completed human calls + initial abandonments + abandonments after human contact. Calls ending with AI and unknown outcomes are excluded from this denominator.
Human contact = completed human calls + abandonments after human contact. Human attention uses known human-routed calls as its denominator. Total abandonment includes both abandonment types and overlaps with human attention; contact can precede abandonment. Post-contact abandonment uses human-contact calls as its denominator.
AI endpoint share uses all inbound calls. An AI endpoint does not prove resolution or exclusive AI handling. Outbound calls are recorded attempts, without verified connection or sales success.
Wait statistics use routing time in known human-routed inbound calls, including zero waits. Seconds are a provisional operational assumption because source units require confirmation. Extreme observations are retained. P90 uses linear interpolation.

SURVEYS
Primary NPS includes valid integer scores 0-10: 9-10 promoters, 7-8 passives and 0-6 detractors. The score is 100 times the difference between promoter and detractor counts divided by all valid responses. Zero inclusion is provisional. The alternative NPS excludes zero from both numerator and denominator and is a sensitivity check, not improvement in customer experience. Response counts for both policies are provided.
Recommendation coverage is valid recommendation responses divided by all inbound calls. It is not a survey invitation response rate.
Primary human-agent satisfaction follows the brief's 1-10 scale and uses known calls with human contact. Zero scores are excluded pending clarification. A 0-10 sensitivity and its response count are also included. Satisfaction coverage uses the known human-contact cohort. Survey respondents may not represent all callers.

COMPARISONS
Periods are consecutive reporting periods relabelled in order. The last period is partial: avoid comparing its absolute volume with complete periods. Team labels refer to initial routing, not final agent or department. All ratios are recomputed from their numerators and denominators, not averaged across periods. Known-outcome shares use all inbound calls; minor unknown outcomes are omitted, so the displayed shares may not add to 100%.

DATA RELEASE POLICY
Only allowlisted numeric aggregates and manually defined English labels are exported. The release excludes individual records, personal names, nicknames, agent identifiers, telephone numbers and mappings, interaction identifiers, calendar dates, timestamps, comments, source reasons, original category strings, source paths, original workbooks, model files, company names and AI-provider names.
No agent-level exports are included. Small classification review categories are merged into Other and unclassified. Nonzero published typology-by-channel cells contain at least 10 records; structural zeroes are retained. Team survey scores are omitted to avoid releasing small respondent cohorts. The Other routing team combines unspecified and unresolved source groups.
The headline aggregates reconcile with the saved and reopened native Power BI model. Observed associations do not establish causes, first-contact resolution, revenue, sales conversion, staffing requirements or service-level compliance.
