311 Toronto Contact Centre: Operations BI

Channel shift, service level, follow-up demand, intraday demand, a 12-month call-volume forecast and an Erlang C staffing model for a contact centre handling about 1.3 million contacts a year, built end to end from public data.

What the data says

Four findings and the operating decision each one points to.

Channel shift

Digital channels are taking share from the phone queue. Self-service (web and mobile) has more than doubled its share since 2022.

Contacts per month, by channel group

All channels. Mail is excluded (under 10 contacts a month).
View as table

Share of contacts, by channel group

Monthly share of all contacts, %. Phone (the rest) is in the chart to the left. The Feb–Mar 2022 counter-visit spike is flagged in the data-quality log.
View as table

Service and follow-up demand

The phone queue has met its 80/75 target three years running, but the margin is thin. Follow-up "where is my request?" contacts peak every summer.

Phone service level vs. target

Share of calls answered within 75 seconds, by year. Target 80%.
View as table

Follow-up contact rate

Status inquiries on an existing request as a share of all contacts, by quarter.
View as table

Forecast and staffing

A damped Holt-Winters model forecasts monthly phone volume 12 months out. It feeds an Erlang C model that sizes the peak-hour team needed to hold 80/75.

Phone contacts per month: actual and 12-month forecast

ActualForecast80% interval
View as table

Staffing what-if (Erlang C)

Agents needed in the peak hour of a forecast month to answer 80% of calls within 75 seconds. Average handle time and shrinkage are planning assumptions, not published data; change them to test the plan.

When and why people contact 311

Demand is a weekday, business-hours pattern. Billing questions lead the general inquiries, which makes them the clearest self-service opportunity.

Average service requests per hour, by weekday and hour

View as table

Top general inquiries

Phone, email, chat and social inquiries about City services.
View as table

Top service-request types, last 12 months

Requests created, all channels.
View as table

Data governance

Every KPI above has a written definition, and every load runs automated quality checks. The raw service-request CSV is unquoted, so about 35,000 rows a year split on a comma inside a division name; the pipeline detects and repairs them, and logs every rejected row.

KPI dictionary

Data-quality checks, latest load