Citation persistence: track exact-URL continuity

By Abiot Y. Derbie - Founder, RankEcho - Updated 2026-09-07

What is the scope of citation persistence?

The unit is an exact target URL within one stable prompt-surface cell observed over ordered post-shipment windows. The method does not ask whether any page on the domain appeared, whether answer prose stayed similar, or whether a page change caused the observation.

Use one predeclared schedule and retain every window. W1 is the anchor-positive post-change cohort definition, not a reconstructed baseline. Do not extend sampling only after a loss, omit an unobserved result, or move a later positive into the original cohort.

How does this differ from citation rate, volatility, and a fixed retest?

Citation rate pools declared positives across eligible cells within a window. Answer volatility compares unordered feature-signature pairs. A fixed-prompt retest compares baseline with a later window. Persistence begins only when W1 contains the exact target citation and preserves the order of every later scheduled window.

These methods can coexist, but their numerators, denominators, order, missingness rules, and questions are not interchangeable.

Which frozen windows define the synthetic record?

The five-edge artifact was recorded as shipped at 2026-08-03T16:00:00Z. W1 through W4 are later scheduled observations; the baseline wrong-page ledger is contextual history and does not enter the persistence denominator.

Every cell retains its original prompt, surface, locale, signed-out synthetic session, exposed-model label, qualifying-link rule, and exact-URL normalization rule.

What does the worked synthetic persistence ledger show?

SYNTHETIC EXAMPLE — Alderwave Relay, its pages, prompts, surfaces, citations, links, and observations are fictional; this is not RankEcho data, customer data, provider evidence, or a benchmark.

T means the exact normalized T01 URL is cited; O records a shift to another owned URL without T01; F and U retain unobserved failed or unavailable states. An observed answer without T01 would remain a separate target-absence state. C01–C04 form the W1 anchor-positive cohort. C05 and C06 are later new gains at W2 and remain outside that cohort.

How does the synthetic persistence arithmetic reconcile?

The primary W1 cohort contains four cells. All four remain confirmed through W2. At W3, C02 has an observed loss and C03 has a failed gap, leaving two confirmed, one lost, and one indeterminate. At W4, C04 becomes unavailable, leaving one confirmed, one lost, and two indeterminate.

C02's W4 target citation does not erase its W3 loss. C03's W4 citation does not fill its failed W3 observation. The partition therefore moves 4/0/0, 4/0/0, 2/1/1, and 1/1/2 across confirmed continuous, confirmed lost, and indeterminate states.

Observed-follow-up denominators remain separate from coverage. Exact T01 is present in 4/4 observed cohort cells at W2, 2/3 at W3, and 3/3 at W4; coverage is 4/4, 3/4, and 3/4. W4 includes one uninterrupted retention, one reacquisition after a shift, and one target observation after a failed gap.

How do observed-follow-up denominators and coverage stay separate?

The exact-target-present column uses only eligible observed W1-cohort follow-ups as its denominator. Coverage uses the full four-cell anchor-positive cohort. Failed, unavailable, and unclassifiable rows reduce coverage and remain unknown rather than entering the target-absence numerator.

Point presence does not repair continuous history: C02 is reacquired at W4 after an observed shift, while C03 shows T after a failed gap. Both are distinct from C01's uninterrupted retention.

Which state rules prevent favorable denominator drift?

Apply status cumulatively from W1 through the stated window; every W1 cohort cell must occupy exactly one status.

How are failures and a zero inception denominator reported?

Failed, unavailable, observed-unclassifiable, and context-drifted windows never become observed losses. They make continuous persistence indeterminate from that point because the planned sequence cannot be confirmed.

If W1 contains zero exact-target citations, report ‘persistence not available: no inception cohort.’ The undefined zero-cohort result must not render as 0%, and later incident citations must not be used to reconstruct W1.

What can RankEcho report and what remains a companion method?

RankEcho can retain configured prompt observations, sources, and shipment-linked later runs under its product definitions. The broader proof-report template can document planned windows, raw evidence, losses, failures, and interpretation limits.

RankEcho does not currently compute or publish this exact-target-URL inception-cohort persistence classification, its absorbing confirmed-loss rule, or its confirmed/lost/indeterminate table. Attach the companion cell-and-window ledger; do not present it as a native product metric or import format.

The Proof Ledger product held rule collapses engines within each run, counts any target citation, and can remain held when at least half of later run rollups are cited, so intermittent uncited runs may still qualify. This manual method instead keeps exact URL identity, consecutive planned windows, reacquisition, and observation gaps explicit.

The Q3 report's citation stability is full cited-domain-set Jaccard overlap across two matched runs. It is not exact-target URL retention in an anchor-positive post-change cohort.

What interpretation is supported?

One of four W1 target-cited cells remains continuously confirmed through W4; one has a confirmed loss and two are indeterminate because of observation gaps. Two additional cells first cite T01 in W2 and remain outside the W1 inception cohort.

These are descriptive post-shipment observations. They do not show that the internal links caused a citation, estimate long-run durability, reveal a provider mechanism, or promise a later citation, recommendation, ranking, referral, conversion, or revenue result.

What deterministic validation completes the persistence record?

Another person must be able to derive the cohort and every cumulative status directly from the complete observation ledger.

When should monitoring continue or stop?

Complete the predeclared schedule regardless of direction. After W4, choose a new prospective schedule only if the decision still warrants it; preserve the original cohort and do not rewrite its endpoint.

A loss can return the team to Find, while an indeterminate result can justify repairing collection. Neither state authorizes selective deletion or a claim that persistence failed across all future windows.

Who maintains this guide?

Written and maintained by Abiot Y. Derbie. Published 2026-09-07; sources checked 2026-09-07; last updated 2026-09-07. The persistence record is synthetic.

Recheck provider source displays, product reporting definitions, cadence, and evidence captures before applying the method. Send material corrections through the contact page.

Sources reviewed

Provider eligibility and measurement claims below were checked against primary documentation. These records do not establish a universal selection formula, causation, or a guaranteed ranking, impression, recommendation, or citation.

Frequently asked questions