Evidence-Driven Salesforce Performance Tuning
Overview
Optimize the end-to-end workload without trading correctness, permission enforcement, shared capacity, or recovery for a synthetic benchmark.
Prerequisites
- Representative workload, latency and throughput objectives, business invariant, and current telemetry
- Queries, object volumes, selectivity, pagination, automation, locks, API mode, limits, and downstream timing
- Authorized performance org or sandbox, synthetic dataset, stop thresholds, and rollback owner
Tool Discipline
Use Read, Glob, and Grep to inspect approved repository and evidence files, WebFetch to re-check current first-party Salesforce documentation, and Write or Edit only for secretless plans, fixtures, configuration, and redacted receipts.
Current Contract
Salesforce offers versioned query and limits resources; query performance feedback through REST explain is explicitly Beta. Query plans, indexes, object cardinality, automation cost, and limits depend on the target org and must be measured.
Authentication
Use an approved read-only performance principal first and a separately approved principal for canary writes. Do not collect broad record payloads or use administrator access to mask real permission cost.
Instructions
- Define the business invariant and measure end-to-end latency, throughput, error rate, API consumption, locks, retries, and downstream lag.
- Trace time across client, network, authentication, query, pagination, automation, async jobs, event delivery, and reconciliation.
- Inspect query shape, selected fields, predicates, cardinality, selectivity, ordering, pagination, and current metadata.
- Use documented explain or query-plan tooling only under its current status and corroborate recommendations with measured tests.
- Compare bounded changes to query design, caching, conditional requests, composite or bulk modes, concurrency, and scheduling.
- Canary one change at a time with synthetic or approved non-production data and fixed stop thresholds.
- Verify correctness, permissions, limits, event and business outcomes; roll back regressions and publish the evidence.
Approval Boundaries
Do not request indexes, change automation, increase concurrency, add caches, alter consistency, or load production for a benchmark without owners.
Output
Return the baseline trace, bottleneck evidence, tested alternatives, capacity effect, correctness checks, canary result, rollback, and recommendation.
Error Handling
| Condition | Response |
|---|---|
| Explain output is treated as production proof | Label it Beta guidance and require representative measured validation. |
| Optimization changes returned records | Reject the change and restore the original business invariant. |
| Faster client behavior consumes unsafe shared capacity | Apply backpressure or scheduling before increasing rollout. |
Example
A redacted completion receipt might look like this:
workload=account-sync; baseline_p95=recorded; bottleneck=query; change=selective-filter; limits=safe; correctness=exact; canary=pass
Resources
Next Steps
Run the workflow first in the lowest-risk authorized org and preserve its redacted receipt. Schedule a review against the next Salesforce seasonal release and the customer change calendar.