Automatic Call Distributor

What an ACD does

When a call leaves the IVR, the ACD takes over. It:

• Places the call in a queue based on the caller's intent, language, customer tier or the number dialed.
• Selects an agent using the routing rules for that queue.
• Plays queue treatment while the caller waits: messages, expected wait time, callback offers.
• Tracks agent states: available, on call, hold, after-call work, break, training and other aux codes.
• Records every event with timestamps, which become service level, speed of answer, abandonment, handle time and adherence reports.

Today most ACDs are part of a cloud contact center as a service platform and route chats, emails, messages and callbacks as well as voice, often in a single blended queue.

Common routing methods

• Longest idle agent: the call goes to the agent who has been available longest. Spreads work evenly and is the default in many centers.
• Fixed order or top-down: agents are tried in the same order every time. Useful for overflow groups but unfair as a main method.
• Round robin: calls rotate across agents in sequence.
• Skills-based routing: each agent has skills (billing, Spanish, technical tier 2) with proficiency levels, and each call carries required skills. The ACD matches them. This is the most common method in medium and large centers.
• Priority routing: high-value customers or urgent call types jump the queue or get a shorter wait target.
• Predictive or attribute-based routing: uses data about the customer and agent outcomes to pick the agent most likely to resolve the call. Powerful, but harder to explain and audit.

Skills-based routing in practice

Skills-based routing lets a center use one pool of agents across many call types instead of separate small teams. Larger pools are more efficient: an Erlang C model shows that one group of 60 agents answers faster than three groups of 20 handling the same total volume.

The trade-off is complexity. Too many skills, or too many proficiency levels, create routing that nobody can predict, and forecasting and scheduling become harder. Practical guidance from experienced planners:

• Keep the number of skills small and tied to real differences in knowledge or language.
• Give new hires one or two skills after training and add more as they pass nesting and quality checks.
• Review skill assignments monthly. Out-of-date skill profiles are the most common cause of misroutes and unexplained queue spikes.
• Use overflow rules with a time delay, for example send billing calls to the general pool after 60 seconds in queue.

ACD data and reporting

Because the ACD time-stamps every call and every state change, it is the source for most operational call center KPIs: calls offered, answered and abandoned, average speed of answer, service level, talk, hold and wrap time, and agent state durations. Workforce management systems pull this data to forecast and to measure schedule adherence.

Check definitions before comparing reports across platforms or sites. Whether short abandons are excluded, whether IVR time counts toward wait, and how transfers are counted all vary by vendor and configuration.

Common ACD setup mistakes

• Routing rules copied from the old system without checking whether they still match the call mix.
• Aux codes nobody defined. If agents pick "Other" for 20% of their time, shrinkage cannot be managed.
• No callback option at peak, so queues build and abandonment rises.
• Skill changes made ad hoc by supervisors without telling workforce planning, which breaks the forecast.
• Ignoring fairness. Routing that always sends the hardest calls to the same strong agents burns them out.