Bot Insights — edge & origin cost

Edge & Origin Cost — Prod, ASN impact queue

Top entities ranked by triggered cache-busting and origin-impact signals for the current window.

Comparison window
2026-04-07 00:00 → 2026-04-14 00:00 UTC
vs 2026-03-31 00:00 → 2026-04-07 00:00 UTC
Executive summary

What this report says

2 of 3 ASNs need analyst attention — start with ASN 64500 (top 2 entities concentrate 64% of origin pressure; top path /api/v1/pricing carries 72% of cache misses).

1 to watch.

Recommended action Investigate origin cost share (affects 1 ASN).

2 Assign 1 Watch 0 Insufficient data 0 Close — expected

2 ASNs need analyst attention; 1 to watch (out of 3).

Edge & origin evidence

3 entities need analyst attention

ASN 64500 #1
Assign Score 78 Origin impact
Edge & origin signals
  • Origin Cost Contribution High +24 pts

    Origin cost contribution is 42.00% of fleet origin pressure.

  • Origin P95 Delta High +16 pts

    Origin p95 latency increased by 310.00%.

  • Cache Miss Rate High +12 pts

    Cache-miss rate is 68.00%.

  • Query String Diversity With High Miss Rate +10 pts

    Query-string diversity is 1240 unique QS with high miss rate.

ASN 64600 #2
Assign Score 50 Cache busting
Edge & origin signals
  • Origin Cost Contribution High +18 pts

    Origin cost contribution is 22.00% of fleet origin pressure.

  • Cache Miss Rate High +12 pts

    Cache-miss rate is 72.00%.

  • Cache Miss Delta High +12 pts

    Cache-miss rate increased by 14 pp week-over-week.

  • Query String Diversity High +8 pts

    Query-string diversity is 890 unique QS.

ASN 64700 #3
Watch Score 14 Cache busting
Edge & origin signals
  • Cache Miss Rate High +12 pts

    Cache-miss rate is 54.00%.

Triage queue

3 ASNs, ordered by what to do

Verdict ASN Score Δ Primary domain Top evidence
Assign
Assign ASN 64500 78 ±0 Origin impact Origin cost contribution is 42.00% of fleet origin pressure.
Assign ASN 64600 50 ±0 Cache busting Origin cost contribution is 22.00% of fleet origin pressure.
Watch
Watch ASN 64700 14 ±0 Cache busting Cache-miss rate is 54.00%.
Top paths

3 path candidates ranked by combined cache-miss and origin pressure

# Path Cache-miss share Origin pressure share Score Band Confidence
1 /api/v1/pricing 72.00% — 88 High review Medium
cache_miss_pct +12pp, origin_p95_ms_pct +37pp, requests_pct +7pp
2 /api/v1/inventory 64.00% — 62 Medium review Medium
cache_miss_pct +8pp, origin_p95_ms_pct +30pp, requests_pct +5pp
3 /api/v1/search 52.00% — 34 Observe Medium
cache_miss_pct +3pp, origin_p95_ms_pct +6pp, requests_pct +4pp
Hosts in scope

Scorecard rollup

Per-host scoring detail lives in the scorecard brief. This rollup only cross-references which hosts the movement applies to.

Host Score Band Verdict
ASN 64500 78 High review Assign
ASN 64600 50 Medium review Assign
ASN 64700 14 Observe Watch
Domain score matrix

How risk points distribute across domains

ASN Score Cache busting Origin impact
ASN 64500 78 22 56
ASN 64600 50 32 18
ASN 64700 14 14 —
Recommended next steps

Investigations to queue from this report

  1. ASN 64500 accounts for 42% of origin requests; confirm whether query-string variance is driving uncacheable traffic and whether a cache-key normalization rule would reduce origin load.
    1 ASN · ASN 64500
  2. Origin p95 jumped 310% this week; check for upstream resource contention or a deployment that changed response characteristics.
    1 ASN · ASN 64500
  3. 890 unique query strings suggest automated request variance bypassing cache; investigate whether cache-key normalization or query-string allow-listing would reduce miss rate from 72%.
    1 ASN · ASN 64600
  4. 22% origin cost contribution combined with ASN 64500 totals 64%; verify whether both ASNs represent the same operator and whether coordinated remediation is appropriate.
    1 ASN · ASN 64600
Coverage

Which rules could be evaluated

Rule coverage by domain — 2 domains evaluated, 10.00% of rule evaluations had missing inputs
Domain Triggered Below threshold Inputs missing Distribution
Cache busting 6 0 1
Origin impact 3 0 0

Confidence: Medium · 3. Reasons: Baseline window has enough rows, Current window has enough rows, Dimensions fit retained schema, Summary table used.

Method & caveats

What this report is and isn't

Rule-based scorecard for Prod · Bot Insights, built from mechanical features only, compared against week over week. It reports what was measured, not why. Missing feature inputs are reported as such — they are not scored as safe.

Schema, source table, and constraints
Schema
bot_scorecard_artifacts.v1
Comparison
Week over week
Producer limit
5 (returned 3, truncated: false)
Tenant / database
Prod · Bot Insights
Table
bi_summary_hour
Constraints
Rule-based scorecard; Mechanical features only; No causal claim; LLM may summarize structured evidence only
Confidence reasons
Baseline window has enough rows; Current window has enough rows; Dimensions fit retained schema; Summary table used
Orientation — what this report measures
What this measures

A cost-impact score for each ranked entity (ASN, host, or bot class) on a 0–100 scale. Higher scores reflect more triggered cache-busting and origin-impact signals — cache-miss rate / delta, query-string diversity, origin p95 delta, origin-cost contribution share.

How to read the score

Higher score = more triggered edge/origin rules. Bands: escalate, monitor, observe.

  • escalate · 0–40
  • monitor · 40–70
  • observe · 70–100
What this can't say

Origin cost is reported as a percentage share, not a billing figure. Missing inputs are reported as missing — they are not scored as zero cost.