2026-08-05

TL;DR

• UK is the whole story again (Δ1d −0.6, and the only market moving — AT/AU/BR/DE/IT all flat-to-up). Top-3 137, top-10 306 are both month-lows; the compression that's rolled through the board since 07-28 is now on its 4th consecutive UK down-day without the rebound that closed the four prior instances.
• But it is still a within-top-20 reshuffle, not a loss. Of 322 terms that were top-10 on 08-04, 319 are still ≤20; only 3 crossed fully out. Ranked count actually rose 340→342, top-20 = ranked. Nothing has left the board.
• The honest caveat: "no visibility lost" is true for ranked count, but UK avg has drifted 3.6→5.5 across the compression (Δ30d −1.3). A city-head sitting at #14 instead of #2 earns a fraction of the clicks even though both count as "ranked." This instance, unlike the prior four, is lasting long enough that the CTR cost is becoming real.
• Ruled out technical (all cliff pages 200/clean today) and competitive (incumbents frozen). Ruled out global algo despite confirmed industry-wide Aug 1-3 volatility — a global event would hit all six markets, and DE/AT/AU are rising.
• Not escalating to red: no net loss, two of yesterday's three stayout-worries recovered, pages healthy. But if UK t3/t10 don't bounce in 1-2 days, treat it as a real click loss and pull GSC.
• IT capture intact (quiet rotation day); BR still compounding (Δ30d +2.0).

What moved and why

UK compression — deepest, longest instance of the month (likely still SERP-local, but no longer costless). The daily band trend since the 08-02 recovery peak tells it plainly:

Four straight days down on t3, and today t10 finally cracked (−16). The four prior instances this month all reversed inside 1-3 days; this one hasn't. Verdict: this is a genuine, sustained local UK SERP recompression — confidence high. What keeps it amber not red: the aggregate carries zero net loss (top-20 = ranked, ranked +2), and the dispersion of the 08-04 top-10 cohort is still overwhelmingly a shuffle — 107 t3 held, 33 t3→4-10, only 3 crossed out of top-20 (coventry, stratford, rochdale), offset by 5 entering.

Today's three out-crossers are the notable part: coventry (2→OUT), stratford (4→OUT), rochdale (4→OUT) — all three had been stable page-1 heads, cliffing clean. That echoes the 08-03 pattern (southend/wolverhampton/hertfordshire cliffs), most of which came back. Encouragingly, yesterday's two stayout-worries both returned: canary wharf OUT→5, northampton OUT→20. Central-london is the one that's now 2 days out (9→OUT→OUT).

IT — quiet rotation, capture holding (confirmed). IT Δ1d −0.3 reads worse than it is: t3 9→6 but top-20 flat (166→167), t10 113. That's the same top-3↔page-1 flapping the EA-vacuum case has described all week (palermo/perugia/firenze bouncing 3↔6-7). The briefing's escort-city →OUT wall — including escort a prato (10→OUT, vol 90,500) — is habitual tail oscillation: prato has flapped 6-11 for a week and blinked out for a single day. Not the capture reversing.

Investigations

• UK cohort dispersion (08-04→08-05): ran the top-10 cohort forward. t3 cohort (144): 107 held / 33→4-10 / 3→11-20 / 1 out. t4-10 cohort (178): 30→t3 / 127 held / 19→11-20 / 2 out. So the t10 −16 is 22 terms sliding into the 11-20 band, and only 3 leaving entirely.
• Technical rule-out: coventry, stratford, rochdale, central-london all snapshot 200 / empty robots / clean self-canonical on today's crawl. Live-fetched /uk/escort/coventry/ (the most surprising, #2→OUT): serving, "Coventry escorts | Erobella UK", 367 profiles, no noindex. Not technical.
• Competitive rule-out: on all three fresh cliffs plus central-london, incumbents are frozen — vivastreet #1 everywhere, kommons/adultwork/ukadultzone/escortrankings in their usual slots. uk.skokka drifted the wrong way overall (9.0→9.6 on 350 kws); its local presence (coventry #6, central-london #17) is a recompression symptom, not a takeover.
• Global-algo rule-out: other markets' top-10 over the same 4 days — AT 384→386, DE 528→539 (rising), IT 117→113, AU 120→123. WebSearch confirms real industry-wide volatility Aug 1-3 (unconfirmed by Google, "YouTube preference"), but a global event can't hit only UK while DE climbs. Coincident, not causal.
• GSC still unavailable in this env — as on prior days, I can't isolate impressions on the compressed heads, which is exactly the number that would settle "reshuffle vs. real click loss." Substituted dispersion; it proves nothing left top-20 but can't price the CTR cost of the within-band drift.

Opportunities

• IT EA-vacuum capture — still the highest-ROI action on the board. Board at period-high depth (ranked 167, t10 113), and the named cities (palermo, perugia, firenze) are flapping 3↔6-7 — an internal-link lift would nail them to #3. Window still decaying toward EA's eventual .net reclaim; bank this week.
• BR still compounding quietly (Δ30d +2.0, avg 8.9). Internal-link window into the thin bairro bench remains open.

Recommendations

1. UK — decision point, not action point (yet). Hold the template (four prior reversals earned that patience). But this is the first instance to run 4 days and erode avg by 1.3 over 30 days, so if t3/t10 don't rebound within 1-2 days, stop calling it costless: pull GSC impressions/clicks for the compressed city-heads (coventry, stratford, rochdale, central-london, and the 11-20 band) to price the CTR loss. That pull needs BJ — no GSC in this env.
2. IT — ship the internal-link push on the rank 4-7 EA-vacated cities (palermo/perugia/firenze first). Highest-ROI action available; window closing.
3. Standing anchor-hygiene (low urgency): point city-head anchors at canonical pages (/cardiff/, /central-london/) not the /cf11/, /london-2/ siblings — same cannibalization signature keeps recurring around the cliff terms.

Watchlist

Updated in watchlist.md: UK compression case updated (08-05: deepest/longest, 4th down-day, decision point — still amber, escalate red if no rebound in 1-2d or the 11-20 band starts crossing out); IT capture and BR growth carried forward; central-london added to the UK stayout watch (day 2).