ApexCloud Retention Ops
Workflow
- Read the prompt and its answer template first. Extract the base URL, date range, month list or quarter, A/R as-of date, account IDs, due dates, requested count/order, allowed enum labels, and required JSON shape.
- Fetch only the allowed ApexCloud Retention Operations API endpoints needed for the requested report. Filter dates inclusively and filter rows locally if an endpoint returns a full year or full portfolio.
- Read references/apexcloud_workflows.md for endpoint fields, joins, and report recipes before doing nontrivial calculations.
- Use scripts/apexcloud_report.py for deterministic normalization when helpful. It uses only the Python standard library and outputs helper JSON for QBR rows, account risk facts, receivables/pipeline summaries, and churn export summaries.
- Return only JSON matching the prompt/template. Do not include notes, citations, markdown, or extra keys.
Core Conventions
- Treat ISO date ranges as inclusive. Treat month ranges as inclusive by
YYYY-MM. - Prefer billing snapshot ARR over account
crm_arror profile ARR whenever the output asks for current ARR, ARR at risk, revenue exposure, or the model check asks whether billing ARR was used. - For support counts and SLA calculations, use clean tickets: exclude duplicates, spam, and cancelled tickets. A clean ticket is SLA-compliant only when both first-response and resolution SLA flags are true.
- For A/R overdue exposure in retention tasks, use the older aging buckets:
61_90 + 90_plus. Do not countcurrent,1_30, or31_60as the retention overdue balance unless the prompt explicitly asks for all aging buckets. - For finance operations reviews that say "older aging buckets", start from rows where
61_90 + 90_plus > 0. - For NPS, ignore retracted responses. Use the latest non-retracted response in the requested period for account-level outputs, and monthly metric rows for monthly QBR outputs unless the prompt instructs otherwise.
- For open expansion pipeline, sum open opportunities whose close dates fall inside the requested period.
- If policy code fields contain pipe-separated enum choices and the prompt does not define a separate policy, use the middle option as the standard evidence-backed policy code.
- Use exact enum labels from the prompt/template. Do not invent enum values.
Helper Script
Run from the skill directory or pass its path explicitly:
python3 scripts/apexcloud_report.py qbr --base-url "$TASK_ENV_BASE_URL" --account-id acct_x --months 2026-04,2026-05,2026-06 --start-date 2026-04-01 --end-date 2026-06-30 --review-due-date 2026-07-22
python3 scripts/apexcloud_report.py risk-facts --base-url "$TASK_ENV_BASE_URL" --account-ids acct_a,acct_b --months 2026-04,2026-05,2026-06 --start-date 2026-04-01 --end-date 2026-06-30 --as-of 2026-06-30 --assessment-date 2026-06-30
python3 scripts/apexcloud_report.py receivables --base-url "$TASK_ENV_BASE_URL" --quarter 2026-Q3 --as-of 2026-09-30 --due-date 2026-10-15 --event-id apex_connect
python3 scripts/apexcloud_report.py churn-summary --base-url "$TASK_ENV_BASE_URL" --candidate-ids acct_a,acct_b
The script output is a working table of facts, not a license to ignore the requested template. Use it to populate and check the exact answer shape.
Report Dispatch
- QBR metrics packet: monthly recognized revenue, clean support ticket counts, clean-ticket SLA compliance, monthly NPS, highlights, source labels, review owner, signoff, and four ordered agenda topics.
- Renewal risk queue or retention action board: account profile, billing snapshot ARR, older A/R balance, clean support health, NPS, product usage trend, renewal window, tenure risk, lifecycle status, and open expansion pipeline.
- Receivables and pipeline operations review: older A/R followups, CRM account linking by legal name or alias, Q-period opportunity summaries, all-region HR context, and named event context.
- Churn validation and outreach ranking: row counts, raw feature count, validation accuracy band, tenure direction, candidate probabilities/ranking, cohort checks, and outreach action mapping.
Final Checks
- Confirm every numeric value has the requested precision: currency 2 decimals, percentages 1 decimal, probabilities 3 decimals, counts and ranks integers.
- Sort in the prompt's required order: top-N by risk/probability for queues, standard retention board order for boards, customer name ascending for receivables followups.
- Validate JSON with
python3 -m json.toolorjson.loadsbefore finalizing.