Intercom

Intercom exposes several object types: conversations, tickets, contacts, companies, admins, teams, and Help Center articles. This integration focuses on the two objects that produce directly attributable performance signals — conversations and tickets — which together represent the full spectrum of customer support activity, from informal real-time exchanges to formal tracked requests. Five KPIs cover conversations and two cover tickets, selected against three criteria: ability to attribute outcomes to an individual agent, resistance to gaming through paired counterbalances, and balance between leading and lagging indicators.

Conversation throughput and quality

Five KPIs cover conversations. They address three distinct management dimensions: volume, speed, and quality. These dimensions interact in ways that make it necessary to read the group as a system rather than tracking any single indicator in isolation.

Volume and responsiveness

Conversations Closed counts the number of conversations an agent brings to a closed state within a period. Average First Reply Time measures the average elapsed time between a conversation being opened and the first reply from the assigned agent. These two indicators establish the baseline of support capacity: how many conversations the agent processes, and how quickly customers receive a first acknowledgment. Conversations Closed is the only throughput measure; Average First Reply Time is the only leading indicator in the conversation block, reflecting the state of the queue before outcomes are produced.

The tension between these two metrics is structurally important. An agent closing a high volume of conversations in a short period may be succeeding, or may be accelerating closures at the expense of resolution quality. Reading Conversations Closed alongside Average First Reply Time provides a first check: a high throughput paired with a slow first reply suggests the agent is responsive within conversations but delayed at the intake stage, pointing to queue management issues rather than execution quality.

Resolution quality: the close time and reopen pair

Average Time to Close measures the average duration from conversation creation to first closure. Reopen Rate measures the proportion of closed conversations that were subsequently reopened by the customer. These two indicators are explicitly designed to counterbalance each other. Average Time to Close carries a gaming risk: an agent under pressure to close conversations faster can do so by marking conversations closed before the customer's issue is genuinely resolved. The reopen event, which requires a deliberate action from the customer side, is outside the agent's control and therefore resistant to this form of gaming. A declining Average Time to Close accompanied by a rising Reopen Rate identifies the deterioration in resolution quality that the speed metric alone would conceal.

Customer satisfaction as an outcome anchor

Average CSAT Score captures the mean satisfaction rating submitted by customers at the end of a conversation, on a scale of one to five. The CSAT score is attributed to the agent who closed the conversation, not necessarily the agent originally assigned. This distinction matters: in teams where conversations are transferred before closure, CSAT reflects the experience of the final interaction, not the entire handling chain. CSAT is the only indicator in this integration where the data point originates entirely from outside the team, making it the most robust quality anchor in the group. Its limitation is coverage: response rates typically fall below thirty percent in support contexts, which means the metric becomes statistically reliable only when the period under review contains a sufficient number of rated conversations per agent.

Ticket resolution efficiency

Tickets in Intercom represent a structurally distinct object from conversations. They are formal support requests with a defined lifecycle — submitted, in progress, waiting on customer, resolved — typically used for complex or multi-step issues that require coordination beyond a single exchange. Teams that use both objects maintain two parallel support flows, each with its own management cadence and expectations.

Tickets Resolved counts the number of tickets an agent brings to resolved state in a period. Average Ticket Resolution Time measures the mean elapsed time from ticket creation to resolution. These two indicators mirror the throughput-versus-efficiency tension present in the conversation block. A high Tickets Resolved count achieved through short resolution times may indicate effective handling, or may indicate that tickets are being marked resolved prematurely to improve the metric. Reading the pair together identifies which scenario is occurring: a rising resolution count accompanied by a stable or declining resolution time suggests genuine capacity improvement; the same rising count accompanied by a subsequent increase in ticket reopening or recurrence would point to premature closure.

Scope and limits of the integration

Intercom records conversation states, timestamps, and ratings, but does not measure the quality of the interaction that occurred within those conversations. Two conversations closed in the same time interval by the same agent may reflect entirely different levels of effort, expertise, or customer benefit; the API data does not distinguish between them. Similarly, a high CSAT score reflects a customer's experience at a moment in time, not a comprehensive assessment of the support relationship.

The conversation and ticket objects only capture interactions that are routed through Intercom. Phone calls, informal messages, or support handled through other channels remain invisible to this integration. The reliability of all seven KPIs depends directly on consistent assignment discipline within the team: conversations and tickets that are handled but not formally assigned to an agent will not appear in that agent's metrics, and the aggregated figures will undercount actual activity proportionally to how frequently this occurs.