TS34Performance and deployment

What experience do your visitors actually have?

Field Core Web Vitals are documented and classified at the 75th percentile.

Prepare evidence

In the app: Evidence import
You provide: Field metrics, device and period
Allow for: Depends on evidence needed

Prepare evidence

Understand

Real experience. Measured over time.

Field Core Web Vitals are documented and classified at the 75th percentile.

According to Google

CrUX aggregates observations over a rolling 28-day window, updated daily. A recent fix does not immediately replace that whole period.

Chrome UX Report (CrUX) API Source accessed on 2026-10-26

Check

LCP, INP, CLS. The same field data.

In TSEP Control

You import evidence for this control. The engine assesses it within the declared scope; missing items remain visible.

Your contribution

Define the scope and expectations before collection. Gather the items below and supply the requested reviews.

With your own tools

In PageSpeed Insights or CrUX, record field data, its period, device and URL or origin level. Retain LCP, INP and CLS with coverage information; missing data is not a pass.

Evidence to retain

LCP ≤ 2.5 s, INP ≤ 200 ms and CLS ≤ 0.1 at the 75th percentile for a Good assessment.

What this result covers

Missing CrUX data does not mean the page is fast.

Act

The failing metric. The affected visitors.

Prioritize the failing LCP, INP or CLS in the observed cohort and date the correction.

Recheck

Dated periods. A 28-day window.

With the CrUX API, data is updated daily over a rolling 28-day window. Date the correction and follow the same cohort: a test the next day is not enough to measure its effect across the entire period.

Your reference points

To go further.

Statuses, sources and method.

Understand the statusesC · NC · NT · NA

The same legend in guides, analysis and demo. Each status depends on the control’s evidence and rules.

C · Verified: The criterion is met within the assessed scope. Keep dated evidence for the next comparison.
NC · To correct: Prioritize the failing LCP, INP or CLS in the observed cohort and date the correction.
NT · Not established: Evidence or confirmations are insufficient to conclude. Supply what is missing; this status does not establish an SEO error.
NA · Not applicable: The control does not apply to the scope under its rules, with a retained justification. Missing evidence alone does not make it non-applicable.

Sources and references2 references

Detailed method and rules3 rules

Control method and criteria

Method
Use an identified field source: CrUX, Search Console or RUM. Record period, device, granularity and population; do not conflate URL and origin.

Acceptance criteria
Source, period, granularity, device and LCP/INP/CLS p75 are recorded and classified against cited thresholds; poor values remain visible and inform a decision.

Non-applicability
Flows explicitly outside CWV scope; missing data alone means NT.

Detailed rules

TS34-A01 · Attributable field cohort

Protocol assessment mode : assisted, with supplied evidence and declarations. App availability is shown on the control card.

Scope
Declared performance scope for this control and qualified reading profile.

Evidence
Original JSON with source/date/hash; plan and supplied review when required, diagnostics and limits under docs/performance-evidence.md.

Method
Identify source, request/returned key, URL/origin, device, period and population under performance-evidence-v1. No implicit scope fallback.

Acceptance criteria
Cohort matches the plan with retained dates/provenance. Explicit lab-for-field substitution fails; a different or old cohort remains inconclusive.

When the result remains inconclusive
Insufficient source/plan/dates/measures; unsupported format/profile, scope mismatch, exceeded bounds or missing review under the contract. Absence never becomes C or NA.

Non-applicability
No automatic atomic exemption; whole-control non-applicability requires its own evidence.

Assumptions
Supplied data/reviews are not authenticated. Plan precedes assessment and lab runs; a historical cohort is not preregistration.

Limitations
C describes documentation, not good performance. No display/ranking, demonstrated cause, automatic human review or global conformity. Replay performs no new measurement.

TS34-A02 · Retained p75 measurements and classes

Protocol assessment mode : automatic. App availability is shown on the control card.

Scope
Declared performance scope for this control and qualified reading profile.

Evidence
Original JSON with source/date/hash; plan and supplied review when required, diagnostics and limits under docs/performance-evidence.md.

Method
Read LCP/INP/CLS p75 in native units and compare exactly with cited thresholds, without relabeling FID/TBT. Retain poor values and unknowns.

Acceptance criteria
All three measures are classified correctly. A bound classification contradicting p75 fails; correctly documented poor performance passes.

Non-applicability
No automatic atomic exemption; whole-control non-applicability requires its own evidence.

Assumptions
Supplied data/reviews are not authenticated. Plan precedes assessment and lab runs; a historical cohort is not preregistration.

Limitations
C describes documentation, not good performance. No display/ranking, demonstrated cause, automatic human review or global conformity. Replay performs no new measurement.

TS34-A03 · Reasoned decision bound to measurements

Protocol assessment mode : manual. App availability is shown on the control card.

Scope
Declared performance scope for this control and qualified reading profile.

Evidence
Original JSON with source/date/hash; plan and supplied review when required, diagnostics and limits under docs/performance-evidence.md.

Method
Retain supplied review, author, provenance, date, source hash and decision for each metric; verify binding without fabricating judgment.

Acceptance criteria
Each metric retains its class, reason and action. A contradictory class fails. Missing/incomplete/unbound review is inconclusive.

Non-applicability
No automatic atomic exemption; whole-control non-applicability requires its own evidence.

Assumptions
Supplied data/reviews are not authenticated. Plan precedes assessment and lab runs; a historical cohort is not preregistration.

Limitations
C describes documentation, not good performance. No display/ranking, demonstrated cause, automatic human review or global conformity. Replay performs no new measurement.

Rule version 0.2.0-rc.1 · Public protocol reference · Versioned source

Put it into practice

← Back to controls