Crisp

Crisp exposes several object types through its API: conversations, messages, contacts, operators, and campaign events. This integration focuses exclusively on conversations, which constitute the primary unit of work for customer support teams and the object through which operator-level performance can be reliably attributed. Five KPIs were retained, selected against three criteria: ability to attribute to an assigned operator, resistance to gaming, and balance between leading and lagging indicators.

Throughput and quality: reading volume against outcome

Conversations Resolved counts the number of conversations an operator brought to resolution during the period. On its own, this indicator measures throughput but says nothing about the quality of that throughput. An agent who closes conversations prematurely to inflate their count will appear productive while generating customer dissatisfaction. CSAT Score provides the counterbalance: it captures the customer's assessment of the interaction and is controlled entirely by the customer, making it structurally resistant to gaming by the agent. Read together, these two indicators distinguish genuine resolution from administrative closure. An operator with high Conversations Resolved and a declining CSAT Score signals a quality-volume trade-off that warrants management attention.

Speed: the relationship between first response and full resolution

Average First Response Time measures the delay between a conversation being created and the operator's first message. Average Resolution Time measures the full elapsed time from conversation creation to resolution. The two indicators occupy different positions in the causal chain and should not be collapsed into a single speed metric. A short Average First Response Time combined with a long Average Resolution Time reveals that operators engage quickly but struggle to close conversations efficiently — a pattern that typically points to complexity in the request, insufficient authority to resolve the issue, or dependency on other teams. The inverse configuration, a long Average First Response Time followed by a short Average Resolution Time, suggests queuing problems upstream of effective execution. Reading both metrics together identifies where in the support cycle delays actually occur, which is not visible from either indicator alone.

Open Backlog as an early-warning signal

Open Backlog counts the conversations currently assigned to an operator and still in an unresolved state. Unlike the other four KPIs, which are calculated over a historical period, Open Backlog is a snapshot indicator. Its value lies in anticipation: a growing backlog at the operator level precedes a degradation of resolution time and first response time before those lagging and leading metrics yet reflect the deterioration. In a support team managing capacity across multiple agents, monitoring individual backlogs allows redistribution of workload before service levels are breached. This makes Open Backlog the primary instrument for short-cycle operational management within this integration.

Scope and limits of the integration

Crisp records conversation states and timestamps but does not capture the substance of what made an interaction effective. A conversation resolved quickly with a high CSAT Score may reflect a trivial request handled well, or a genuinely complex problem resolved through exceptional expertise; the data cannot distinguish between the two. CSAT data is collected only when the customer chooses to leave a rating, which typically covers between ten and thirty percent of conversations. For individual operators with low conversation volumes in a given period, the CSAT Score is derived from a sample too small for statistical confidence and should be interpreted with caution.

Attribution relies on the current assignee field at the time of data extraction. Conversations that were reassigned during their lifecycle are attributed in full to the final operator, with no record of prior assignment visible through the standard list endpoint. The reliability of all five KPIs depends directly on assignment discipline within the team: conversations left unassigned, or assigned to a generic shared account, will not surface in operator-level reporting. Consistent use of assignment is a prerequisite for this integration to produce meaningful data.