gabrielmoreira Clari Cost Tuning
Overview
Minimize Clari API overhead: reduce export frequency, cache aggressively, export only needed data types, and monitor usage.
Prerequisites
- An approved consumer inventory and freshness requirement for each report
- Usage telemetry for API calls, export jobs, warehouse loads, and cache age
- Data-owner approval for retention, aggregation, and delivery changes
- A non-production validation path for schedule or payload changes
Instructions
Export Only What You Need
# Full export (6 data types) -- more API load
full_types = ["forecast", "quota", "forecast_updated",
"adjustment", "crm_total", "crm_closed"]
# Minimal export (2 data types) -- faster and lighter
minimal_types = ["forecast", "crm_closed"]
# Use minimal for dashboards, full for audit/compliance
Optimize Export Frequency
| Use Case | Recommended Frequency |
|---|---|
| Executive dashboard | Daily |
| Forecast accuracy tracking | Weekly |
| Compliance audit | Quarterly |
| Ad-hoc analysis | On demand |
Cache to Avoid Redundant Exports
# Cache recent exports (see clari-performance-tuning)
cache = ExportCache(ttl_hours=8)
def smart_export(client, forecast_name, period):
cached = cache.get(forecast_name, period)
if cached:
print(f"Cache hit for {period}")
return cached
data = client.export_and_download(forecast_name, period)
entries = data.get("entries", [])
cache.set(forecast_name, period, entries)
return entries
Usage Tracking
class ClariUsageTracker:
def __init__(self):
self.api_calls = 0
self.exports = 0
def track_call(self):
self.api_calls += 1
def track_export(self):
self.exports += 1
def report(self) -> dict:
return {
"api_calls": self.api_calls,
"exports": self.exports,
}
Error Handling
| Condition | Response |
|---|---|
| Consumer needs fresher data than the approved schedule | Obtain owner approval and measure incremental load before changing cadence. |
| Cache returns an expired or mismatched period | Discard it and run the controlled export path. |
| Usage spikes or provider throttles | Reduce scope/concurrency and preserve the last certified dataset. |
| Cost reduction removes an audit-required field | Reject the change until the compliance owner approves an alternative. |
Output
Publish a cost-and-usage decision with current and proposed cadence, payload scope, cache policy, observed API/warehouse load, forecast impact, owner, and rollback threshold. Keep business amounts and rep-level data out of the operational report unless the recipient is authorized.
Examples
Move a non-critical dashboard from hourly full exports to daily minimal exports in staging, verify freshness and audit coverage, and compare job count and warehouse cost for a week. Roll back the schedule if the certified data window or required field coverage no longer meets the report contract.
Resources
Next Steps
For architecture patterns, see clari-reference-architecture.