Monthly Sales Performance Analysis

Sales Performance Analysis produces one leadership report about what helps win and grow accounts, what costs business, and what the team should repeat or change. It covers new business and sales to existing customers, including expansion, renewal and repeat purchases. Individual deals provide evidence for team-wide findings.

Set up the report

Open Background Agents and choose Sales Performance Analysis. The builder inspects available CRM context, then confirms your priorities, sales scope, monthly schedule and report recipients. Check the displayed time and timezone. Confirming the setup creates the agent; there is no separate publish step in the builder.

This is one scheduled workspace-wide run, not a separate run for every deal. Put the relevant teams, pipelines, sales types and exclusions in the brief. Do not configure per-deal event triggers for this template. Creating or changing the agent requires the authoring capability described in Agent permissions.

Historical context and evidence

Monthly describes when the report arrives, not how far back it can look. The investigation chooses and extends its history using sales cycles, available outcomes and changes in pricing, customer mix or sales approach. It can compare earlier and recent performance, including an earlier approach that worked better. It does not need to reread years of every conversation.

Population metrics require the relevant accessible deal population. Detailed research into selected cases can explain or challenge patterns, but cannot establish a company-wide win rate or prove that a change caused better results. The report should distinguish these scopes and include original-source links, counterexamples and material access limits.

The lead investigates connected CRM, notes and meeting evidence, and can read connected email and calendars for active members of the same customer. It checks connection coverage and keeps account provenance when the same conversation appears in multiple mailboxes. Disconnected or failed accounts and missing transcripts remain explicit access gaps. The lead uses a private sales-analysis skill to read conversations in order, separate buyer statements from seller claims, and test patterns against counterexamples. Public research can add context; private conversations and commercial terms must stay out of public search queries.

Results and delivery

The report is saved in the agent's history and delivered only through the configured channels. Email always includes the agent owner and any additional members selected during setup; owner-excluded email delivery is unavailable. Confirm this full set. Slack sending must be enabled in agent settings before Slack delivery can occur. Research does not update CRM records, create tasks or send outreach.

Each report should be useful without a follow-up conversation. It prioritizes supported commercial findings, proportionate decisions or experiments, and how to judge their results. When checked evidence supports no material change, the next report can be a short update with links to still-supported findings. Missing access or insufficient evidence is a limitation, not proof that nothing changed. Prior reports guide later research but do not replace originals.

Credits and incomplete runs

This template uses Usage-based credits on both pricing versions. One report can have several usage rows for research and report generation; group them by run. Cost depends on the evidence read and work performed, not a fixed charge multiplied by the number of deals. Do not promise a fixed cost or complete coverage from deal count alone.

Long research histories are condensed into observational memory, preserving source references, queries, page cursors, partial calculations and unread evidence so retrieval can continue. Reaching the compaction threshold does not end research or establish an access gap. Observation calls use credits and their completed turns survive restarts alongside the research turns.

Research has no separate token, dollar, elapsed-time or total-turn ceiling. Account credits, cancellation or a failed request can stop the investigation. An incomplete draft is not a verified final report, and incomplete research is not delivered. Successfully charged work can remain charged when later work fails. If delivery fails after the report is saved, the report remains available and the run records the failure. Inspect its status and any completed delivery before rerunning, to avoid repeating work or sending another copy. Completed research turns survive service restarts. If an interrupted model request or delivery has an unknown outcome, the run stops for inspection instead of repeating that request automatically.

See Read and manage agent runs and credit estimates and deductions.