Lesson 3 of 6
Volume checks: planning peak-load tests with IT
9 min
Listen to this lesson
A warehouse runs on minutes. If wave release takes 15 minutes instead of 2, or each RF screen takes 3 seconds instead of under one, a shift loses hours, and the 16:00 truck leaves late. Functional tests with five deliveries will never show this. Volume checks do, but only if the business defines what "fast enough" means and IT designs a test that recreates the real peak.
Define the peak in business terms
Start with numbers Marcus recognizes, not technical metrics. Contoso's peak is the last working day of the quarter, when fleet customers order before budget closes:
| Measure | Normal day | Peak day (planning basis) |
|---|---|---|
| Outbound delivery items | 600 | 1,200 |
| Deliveries | 150 | 300 |
| Inbound trucks | 12 | 25 |
| Concurrent RF users | 30 | 45 |
| Wave releases | 4 | 6, the largest with about 400 items |
| Cycle count documents | 40 | 0 (suspended on peak days) |
Contoso's warehouse 1710 volumes. Figures are the project's planning assumptions.
Agree thresholds the business can accept
Thresholds turn "it feels slow" into pass or fail. Each one should come from a business consequence: how long can the PM wave take to release before pickers stand idle? How long can an RF screen take before scanners queue? Sam and Priya propose numbers; Marcus accepts them, because they live with the result.
| Step | Threshold (Contoso) | Business reason |
|---|---|---|
| RF screen response (scan to next screen) | 95% under 1 second, none over 3 seconds | Pickers scan hundreds of times per shift |
| Release of the largest wave (about 400 items) | Under 3 minutes to all warehouse orders created | Pickers wait for work after release |
| Goods issue for one delivery | Under 10 seconds | Trucks queue at doors |
| Planned replenishment night job | Finishes before 05:00 | First wave starts at 06:00 |
| Physical inventory clearing job | Finishes within its batch window | Does not block morning work |
Performance thresholds Marcus signed off before testing.
Design the test with IT
- Volume dataLoad realistic stock, bins and open orders into the test system (similar occupancy to production)
- Background loadRun the night jobs and interfaces that will run in production at the same time
- Generate demandCreate peak volumes of deliveries and inbound deliveries
- Release and executeRelease the peak waves; real or scripted RF users confirm tasks in parallel
- MeasureResponse times, job runtimes and warehouse order creation logs
- CompareAgainst the thresholds; analyze anything that misses
Some practical points. A volume test needs a system sized like production; Sam confirms with the RISE team which system can carry it. Warehouse order creation and wave release are where EWM does heavy work, so activate the warehouse order creation log (/SCWM/ACTLOG) for analysis, and remember to switch logging off afterwards because logs add load. Use the warehouse management monitor and EWM's measurement services or the Warehouse KPIs app to watch open tasks and backlogs during the run. Automated RF scripts can generate load, but at least some real users on real devices should run in parallel, because they feel what the numbers mean.
- Week −10Thresholds agreedMarcus signs the peak definition and thresholds
- Week −8Test design and dataSam and Priya prepare data, scripts and monitoring
- Week −6First volume runBaseline; defects raised for misses
- Week −4Retest after fixesSecond run with fixes and tuned settings
- Week −2Final confirmationRun during the dress rehearsal window if possible
What would you do?
The first volume run releases the 400-item wave in 11 minutes against a 3-minute threshold. Everything else passes. What is the best next step?
| Measure | Threshold | Run 1 | Run 2 (after fixes) |
|---|---|---|---|
| Largest wave release (about 400 items) | Under 3 min | 11 min | 2 min 20 s |
| RF screen response, 95th percentile | Under 1 s | 0.8 s | 0.7 s |
| Goods issue per delivery | Under 10 s | 6 s | 5 s |
| Planned replenishment night job | Done before 05:00 | 04:10 | 03:55 |
Illustrative course data: Contoso's volume runs against the signed thresholds. The wave fix was an index on a custom table read in the warehouse order creation enhancement.
- Define the peak in business numbers that the process owner recognizes.
- Agree thresholds based on business consequences before testing.
- Recreate production conditions: data volume, background jobs, interfaces and parallel RF users.
- Analyze misses with logs and custom code review, fix, and rerun.
Check your understanding · not graded
Lesson check
1. Who should accept the warehouse performance thresholds?
2. Why run night jobs and interfaces during a volume test?
3. Which log helps analyze slow warehouse order creation?
4. Why switch logging off after the volume test?