Tested at up to five million time entries. We tested Benedict App with four synthetic datasets of about 100 thousand, 500 thousand, 1 million and 5 million time entries, and with up to 100 simulated users working at once.
- Everyday work stays fast as data grows. A team member's month view took a median of 6 ms at about 100 thousand entries and 10 ms at about 5 million.
- Under load, with 10–100 simulated users on the largest dataset, half of all requests completed within 19–44 ms and 95 % within 213–326 ms.
- Every financial figure was checked. A separate verification procedure recomputed all 5,835,812 report lines in exact decimal arithmetic, across five currencies. It found zero differences.
- No acknowledged work was lost. All 6,285 time entries saved during the load tests were present exactly once afterwards.
- Month-end close ran under load with no errors: 50 users, the scheduled report generator and full-month exports, all at once.
Where the limits are. Year-long, organization-wide administrator views take seconds at this scale, about 16 s at 5 million entries. Full-year text searches across the whole organization can reach the request time limit at 5 million entries. The full report documents both, together with the available mitigations.
On Azure, Benedict Corp measured release 2.21 with up to 75 simultaneous simulated users and release 2.26 with up to 20, with synthetic data and no errors (see the Benedict App Technical Review).
All measurements above were taken in a local, isolated test environment (SQL Server on a single workstation), with synthetic data and simulated users. They were not taken on Microsoft Azure and are not a service-level commitment. Benedict App is installed in each customer's own Azure subscription. Performance there depends on the service tiers the customer chooses; sizing guidance is supplied with the software.