# Workload chart descriptions

Synthetic workload exploration
Input rows: 25
Distinct observations: 24
Exact duplicates removed: 1
Labels normalized: 1
Total requests: 4180
Observed latency values: 23
Unavailable latency values: 1
Mean daily service p95 (ms): 38.7
Median daily service p95 (ms): 34.0
IQR investigation flags retained: 1
Descriptive synthetic evidence; no causal, production, or model-performance claim.


## latency distribution

Histogram of observed per-service daily p95 latency in milliseconds; counts and interpretation follow.

Observed daily service p95 values: 23; unavailable values: 1. Mean: 38.70 ms; median: 34.00 ms. The companion bin table gives every count. Bins include their lower edge; only the final bin includes its upper edge. Daily service percentiles cannot be pooled into an overall request-level p95.

Underlying values: latency_distribution.csv

## service comparison

Bar comparison of total requests by service, starting at zero; exact totals follow.

authentication_gateway: 1860 requests; network_monitor: 2320 requests. These are totals over supplied observations, not request rates or production estimates.

Underlying values: service_comparison.csv

## daily requests

Line plot of daily combined requests, with missing dates left as gaps; exact date totals follow.

Across 12 supplied dates, total requests are 4180. The first date has 220 requests and the last has 440. The largest daily total is 560 on 2027-01-07. Dates with fewer than two service observations: 0. Gaps between dates break the plotted line; absent dates are not zero. Use the daily table to inspect changes rather than infer a stable trend from a tiny fictional sample.

Underlying values: daily_requests.csv

## request latency

Scatter plot of requests and daily p95 latency by service; complete pairs and interpretation follow.

The relationship plot uses 23 complete request/latency pairs; 1 unavailable latency values are omitted only here and in the distribution. IQR investigation flags: 1; values remain included. Marker shapes and colors both identify services. Some marks may overlap; the paired table preserves every observation and gives exact values. The constructed association does not establish that request volume causes latency.

Underlying values: request_latency.csv

## axis comparison

Two bar plots of the same service totals: a zero baseline and an explicitly truncated-axis teaching example.

authentication_gateway: 1860 requests; network_monitor: 2320 requests. The left panel starts at zero. The cropped baseline is 1767 requests. Visible heights for the smallest/largest totals become 93 and 553, a ratio of 5.95, while the actual total ratio is 1.25. Cropping a bar baseline exaggerates apparent height differences and is deliberately misleading for comparing lengths. The underlying values and their ratios do not change. Use the zero-baseline panel for the comparison.

Underlying values: axis_comparison.csv