Help Scout

Help Scout exposes a range of object types: conversations, threads, customers, ratings, users, and mailboxes. This integration focuses exclusively on conversations, which are the fundamental unit of work in a support operation and the only object type where individual agent attribution is direct and reliable. Ratings are folded into the conversation object through the happiness score. Six KPIs were retained from a longlist of fourteen candidates, selected against three criteria: ability to attribute to an individual agent by email, resistance to gaming or artificial inflation, and balance between leading indicators that anticipate quality and lagging indicators that confirm outcomes.

Volume and quality: reading output against its consequences

Conversations Resolved counts the number of conversations closed by an agent during a period. This is the primary output metric for a support team: it measures how much work moved to completion, not how much work was touched. Happiness Score captures the net customer satisfaction signal attributable to the same agent — the difference between positive and negative ratings divided by total rated conversations.

These two indicators are intentionally paired. Conversations Resolved alone creates an incentive to close conversations prematurely or with insufficient answers — an agent under a resolution-volume target will find it rational to close ambiguous conversations rather than invest in a complete response. Happiness Score neutralizes this incentive: the customer assigns the rating independently after the conversation ends, and a pattern of rushed closures will surface as a degraded score within one to two reporting periods. Neither metric is meaningful in isolation; together, they define the boundary between productive output and output optimized against the metric.

The Reopened Rate extends this quality signal. It measures the proportion of closed conversations that were subsequently reopened by the customer — an event that occurs when the original resolution was incomplete or incorrect. A rising Reopened Rate in an agent with high Conversations Resolved is a precise diagnostic: it signals that closure volume is being driven by premature or superficial responses. This combination of three indicators covers the full output-quality arc and eliminates the blind spots each creates when read alone.

Speed and efficiency: the leading indicators of support experience

Average First Response Time measures the delay between a conversation being created and the agent sending the first reply. It is a leading indicator because it directly precedes and predicts customer satisfaction: empirical data across support platforms consistently shows first response time as the variable most correlated with positive ratings. An agent can act on this indicator daily by prioritizing newly assigned conversations before engaging in deeper work on older threads.

Average Resolution Time measures the full duration from conversation creation to closure. It is a lagging indicator that complements First Response Time by capturing what happens after the initial reply. The relationship between the two is diagnostic: a short First Response Time combined with a long Average Resolution Time indicates that the agent engages promptly but requires many exchanges to resolve — pointing to complexity, knowledge gaps, or unclear communication. Conversely, a long First Response Time combined with a short Resolution Time after engagement can suggest the agent is batching work, which affects customer experience even if the total resolution effort is efficient.

Average Replies per Resolution captures the average number of agent messages sent before a conversation closes. Fewer replies for the same resolution outcome indicates higher communication precision: the agent understood the issue quickly and answered it completely. This indicator carries a moderate gaming risk — an agent could send a single very long reply to minimize the count without improving quality. For this reason, it is read in conjunction with Happiness Score: a low reply count paired with a degraded score signals quantity-gaming rather than genuine efficiency. When the two move together favorably, Average Replies per Resolution becomes the clearest proxy available for communication quality, which no other indicator in this set directly measures.

Scope and limits of the integration

Help Scout records conversation events and customer ratings, but does not measure the substance of what is communicated. An agent can achieve a favorable Happiness Score through empathetic language in a partial resolution, or an unfavorable score through accurate but unwelcome information. The API data does not distinguish between these cases. Similarly, a low Average Replies per Resolution may reflect genuine expertise or the systematic use of scripted responses that do not address the customer's actual question — a distinction invisible to any quantitative indicator.

The Happiness Score is only meaningful above a minimum volume of rated conversations. Help Scout customers rate a subset of conversations, and agents handling low-volume or specialized queues may have too few ratings in a given period for the score to be statistically reliable. The integration does not suppress low-volume scores by default; managers should apply judgment when interpreting results for agents with fewer than ten rated conversations in the reporting period.

All six indicators depend on consistent use of the assignment workflow in Help Scout. If conversations are left unassigned, resolved by a different agent than the one listed as assignee, or closed through automated rules without human review, the attribution becomes inaccurate and the KPI values misleading. The reliability of this integration is a direct function of the team's discipline in maintaining correct assignment at every stage of the conversation lifecycle.