Performance Baseline

This page documents the latency and success-rate baseline for the Public API, captured over a 14-day window. Use these numbers as the reference point when evaluating performance test results or making scaling decisions.

---

Methodology

Data was collected with the following KQL query against Application Insights:

AppRequests
| where TimeGenerated > ago(14d)
| where AppRoleName == "public-api"
// 1. Exclude Health/Liveness checks
| where OperationName !contains "health"
    and OperationName !contains "alive"
// 2. Exclude Browser/Dev noise
| where OperationName !contains "browserLink"
    and OperationName !contains "aspnetcore-browser-refresh"
    and OperationName !contains "robots.txt"
// 3. Exclude Documentation/Root noise
| where OperationName !contains "swagger"
    and OperationName !contains "openapi"
    and OperationName != "GET /"
// 4. Ensure we only look at versioned API calls or specific functional paths
| where OperationName contains "/v" or OperationName contains "api"
| summarize
    RequestCount = count(),
    percentiles(DurationMs, 50, 95, 99),
    AvgDuration_ms = avg(DurationMs),
    SuccessRate = avg(iff(Success == true or ResultCode == "404" or ResultCode == "400" or ResultCode == "401", 100.0, 0.0)) // treat 'Not Found', 'Not Authenticated' and 'Bad Request' as success
  by OperationName
| project
    Endpoint = OperationName,
    Requests = RequestCount,
    P50_ms = round(percentile_DurationMs_50, 2),
    P95_ms = round(percentile_DurationMs_95, 2),
    P99_ms = round(percentile_DurationMs_99, 2),
    Avg_ms = round(AvgDuration_ms, 2),
    SuccessRate = round(SuccessRate, 2)
| order by Requests desc

Success-rate definition: HTTP 400 Bad Request, HTTP 401 Unauthorized, and HTTP 404 Not Found are counted as successful. These are valid, expected API responses (invalid caller input, unauthenticated caller, or a missing resource) and do not indicate a platform fault. Only 5xx responses and network-level failures reduce the success rate.

Captured: March 12, 2026 (covering the preceding 14 days)

---

Results

Individual image proxy requests (GET /v1/items/images/proxy/wiser2/...) are excluded from this table — the query returns hundreds of single-occurrence rows for specific image paths. The aggregated template route GET items/images/proxy/{sequence:int} (4 721 requests) represents this traffic as a whole.

Endpoint	Requests	P50 (ms)	P95 (ms)	P99 (ms)	Avg (ms)	Success %
GET items/price	91 496	92.84	336.08	697.86	126.08	100
GET items/restock-orders	47 188	13.82	25.44	87.11	15.34	100
GET items/attachments	45 983	8.10	13.70	67.63	9.25	100
GET customers/{customerCode}/addresses	43 152	11.45	20.09	69.60	12.58	100
GET pricelists/items/detail	39 515	11.69	32.76	85.47	13.51	100
GET items/images	36 966	10.06	15.67	67.37	11.16	100
GET channels/{channelCode}/items	23 304	75.64	300.87	693.04	109.33	100
GET channels/{channelCode}/items/detail	19 767	34.81	263.04	481.98	72.76	100
GET customers/{customerCode}/contact-persons	19 075	7.88	16.48	76.56	11.18	100
GET items/images/proxy/	4 721	88.70	509.03	1 201.75	170.87	100
PUT items	3 058	40.31	74.17	299.88	48.49	100
PUT customers/{customerCode}/addresses/	1 857	32.64	95.49	1 232.85	62.42	100
PUT pricelists/items	1 102	28.25	104.02	212.19	50.80	100
GET administration/notifications	857	10.50	26.69	121.67	16.00	100
POST items/list	471	1 211.08	7 984.65	24 150.55	2 215.84	100
GET pricelists/items	436	24.48	271.86	537.70	67.71	100
GET pricelists/detail	372	8.38	11.16	31.36	7.99	100
PUT pricelists	342	10.99	14.51	28.33	10.56	100
POST customers/list	318	37.44	133.03	220.71	52.99	100
GET vat-tariffs/rates	305	9.00	73.17	92.44	16.52	100
POST pricelists/items	280	23.21	33.62	49.28	18.44	100
POST items/attachments	268	76.02	132.38	186.10	78.05	100
GET customers/{customerCode}/addresses/	262	11.72	101.65	209.61	24.95	100
DELETE items/attachments	252	36.55	52.22	109.20	38.84	100
GET customers/{customerCode}/contact-persons/	191	11.66	58.19	229.35	19.59	100
GET channels/	180	13.53	87.74	178.05	25.37	100
GET customers/	148	9.45	72.03	203.16	17.82	100
GET webhooks	143	25.96	194.15	194.15	54.85	100
GET auth/me	130	0.57	7.84	192.25	5.70	100
GET attribute-definitions	111	11.93	93.78	211.69	22.84	100
GET pricelists	108	10.56	15.24	298.36	25.21	100
GET items/audit-trails	95	153.67	610.74	990.57	230.23	100
POST items	73	48.67	1 727.72	2 618.62	187.49	100
GET items/detail	72	17.77	175.98	802.96	43.46	100
DELETE customers/	67	29.56	80.32	108.79	35.36	100
GET unit-of-measures	62	11.50	22.61	131.71	14.60	100
GET webhooks/	32	11.31	62.04	81.00	16.70	100
POST items/restock-orders	26	9.61	14.44	81.56	11.82	100
DELETE items	13	25.63	193.81	193.81	40.41	100
POST customers/{customerCode}/addresses	12	48.62	292.01	292.01	80.66	100
GET entitymetrics/total	10	7.23	99.84	99.84	18.70	100
GET channels	10	13.59	219.00	219.00	43.06	100
GET entity-lists	9	10.00	54.56	54.56	15.18	100
GET entitymetrics/counters	8	6.85	89.63	89.63	18.86	100
GET jobs	7	10.04	76.94	76.94	21.70	100
PUT customers/{customerCode}/contact-persons/	5	417.08	2 399.81	2 399.81	875.26	100
GET items/customer-prices	4	14.82	676.76	676.76	178.71	100
GET customers/{customerCode}/audit-trails	4	68.92	188.49	188.49	102.46	100
POST items/images	4	127.19	197.48	197.48	138.65	100
GET channels/{channelCode}/audit-trails	3	75.33	143.16	143.16	97.21	100
DELETE items/images	3	25.25	27.24	27.24	25.45	100
POST customers	3	72.96	110.66	110.66	76.56	100
POST customers/{customerCode}/contact-persons	3	1 304.77	1 404.61	1 404.61	935.40	100
PUT customers/	3	49.26	81.46	81.46	57.48	100
GET flows/definitions	2	4.50	7.12	7.12	5.81	100
GET entity-lists/	1	9.65	9.65	9.65	9.65	100
GET orders	1	6.50	6.50	6.50	6.50	100
DELETE customers/{customerCode}/addresses/	1	32.54	32.54	32.54	32.54	100
GET connectors	1	9.14	9.14	9.14	9.14	100

---

Observations

High-traffic endpoints

The following endpoints account for the vast majority of request volume and deserve the closest attention in performance testing and scaling decisions:

Endpoint	Requests	Notes
GET items/price	91 496	Highest volume; P95 337 ms — monitor closely.
GET items/restock-orders	47 222	Fast; well within SLO.
GET items/attachments	45 983	Fast; well within SLO.
GET customers/{customerCode}/addresses	43 152	Fast; well within SLO.
GET pricelists/items/detail	39 515	Fast; well within SLO.
GET items/images	36 966	Fast; well within SLO.
GET channels/{channelCode}/items	23 304	P95 301 ms; keep under observation.
GET channels/{channelCode}/items/detail	19 767	P95 263 ms; 100% success once 400/404 are excluded.
GET items/images/proxy/	4 721	Image proxy; P95 509 ms, P99 1 202 ms — see latency outliers.

---

Latency outliers

Endpoints with elevated P95 or P99 values that should be investigated or load-tested before raising traffic:

Endpoint	Requests	P95 (ms)	P99 (ms)	Notes
POST items/list	471	7 985	24 151	Batch endpoint — expected to be slow; ensure callers use appropriate timeouts and this does not run at high concurrency.
POST items	73	1 728	2 619	High P99; likely involves multi-step writes. Profile under load.
PUT customers/{customerCode}/contact-persons/	5	2 400	2 400	Very few samples; P95 may not be representative, but worth watching.
POST customers/{customerCode}/contact-persons	3	1 405	1 405	Very few samples; same caveat.
GET items/images/proxy/	4 721	509	1 202	Image proxy at meaningful volume; P99 > 1 s warrants profiling if traffic grows.
GET items/audit-trails	95	611	991	Read-heavy; consider caching or pagination if volume grows.
PUT customers/{customerCode}/addresses/	1 857	96	1 233	P99 spike (1 233 ms) against a P95 of 96 ms indicates occasional slow outliers — likely lock contention or downstream delays.
GET items/price	91 496	337	699	Highest-volume endpoint; P95 is acceptable today but should be a key SLO signal.

---

Using this baseline

• Performance tests — Use the P50, P95, and P99 values above as pass/fail thresholds. A regression is any endpoint whose P95 exceeds the baseline P99 under equivalent load.
• Scaling decisions — Cross-reference with the Scaling page. High-volume endpoints that approach their P99 under increases in maxReplicas or reduced concurrentRequests should be re-profiled.
• Refresh cadence — Re-run the KQL query after significant traffic changes, major releases, or infrastructure resizing events, and update this page with the new snapshot.