Week of Jul 7 to Jul 13. What shipped, then what needs a call. Every metric below is cross-checked against two sources, and the conversion number carries a deliberate asterisk (explained under the table).
| Metric | Value | Read |
|---|---|---|
| Traffic held its gains | 6,259 sessions · flat wk/wk (+1%) | Last week's +24% jump didn't give back. We're holding the best level in months, not spiking-and-fading. (bowtie.) |
| Search visibility | 1,398 clicks · 177K impressions (7-day) | Homepage still the workhorse (573 clicks). (Search Console.) |
| Free-VPN cluster is the conversion engine | 27% on /vpn/best-free-vpn, higher when translated | English free-VPN page converts 1-in-4. Translated versions convert far harder: Persian Windscribe 56%, Arabic free-VPN 39%, Chinese free-VPN 22%. IPTV 21%. (bowtie + /out tracking.) |
| Nord is still the #1 money page | /vpn/nordvpn — 39 conversions, top absolute | The rebuilt provider path is doing its job. (bowtie.) |
| Domain broker converts on intent | /domain-broker 12.4%, get-started 6.5% | Real buying intent — the opposite of the navigational VPN traffic. (bowtie.) |
⚠️ On the site-wide conversion number: bowtie shows 386 conversions at 6.2%, ~42% under the 4-week average. This is a measurement artifact, not a revenue drop — same pattern as last week. The exit-popup (exit-3) removal and the verdict-gateway change deliberately reshaped what gets counted as a conversion (bowtie's own product log flags both with "do NOT read as a drop"). Money-path volume (D1
/outclicks + Nord's report) is the truth source; the per-page rates above are the honest read.
The use-case → money-path attribution is live, and the first ground-truth read is strong: visitors who pick a use-case and reach the /out gateway proceed to the offer far better than those who don't.
| Gateway → send-off (continue rate) | Rate |
|---|---|
| Intent use-cases (enterprise, gaming, streaming, security) | 48–69% |
| No use-case (baseline) | 32% |
/out clicks per visitor vs 1.6 for no-use-case./vpn/best-free-vpn is one of our biggest — while the homepage sets none. So the use-case pool naturally skews free → Proton (its champion). That's the lander structure working, not anything to fix on the Nord path; the two segments simply enter from different doors.The use-case a visitor picks (streaming, gaming, privacy, free…) now follows them end-to-end.
/vpn/, and niche landers all set/read the use-case, so it's captured wherever a cold visitor lands, not just the homepage./vpn/ and the provider pages, then carries all the way to the /out gateway — the interstitial reflects what they came for.sub5 — we can now see which use-case converts best and worst in Nord's own reporting, not just guess.vpn.section_view + vpn.usecase events, and every affiliate_click now carries use_case + placement (which section fired it). So we can slice conversions by what the visitor was shopping for AND where they clicked, without waiting on Nord's report.aff_sub4. The Free section now leads with the honest downsides + a paid-is-better nudge./vpn/ hub makeover (shipped to prod, tag prod-2026-07-10-001)Our #2 landing page got a conversion-first rebuild.
30475d88)The /out interstitial (the money path itself) now speaks the visitor's language across all 17 locales, price chrome and guarantee included. A pre-deploy i18n gate blocks any deploy with missing/untranslated/mis-wired gateway strings. This directly serves our fastest-growing segment — international traffic converts hardest (see the table).
Config-driven partner.geo_offers in PARTNERS_KV, editable in cf-admin (merge-on-save so it won't clobber other partner fields), with a runbook (GEO-OFFERS.md). This is the mechanism to fix last week's "Arabic/China Nord = $0" problem — we can now point those markets at offers that actually fit them.
Definition block, internal-linking overhaul, intent-first intros, duplicate Article JSON-LD merged via shared @id, "premium domain broker" consolidated on the parent, and the recovery cluster re-translated across 16 languages. Feeds the Jul 10 checkpoint decision below.
The /out gateway now preconnects to both the affiliate redirect origin and the provider's actual brand site (Nord, Surfshark, Proton, etc.), warming the final landing so the post-click paint is faster. Tracking untouched. This was the "prefetch Nord on /out" item from last meeting — done.
Button.astro — standardized primary/secondary buttons with states + a real focus-ring (fixes a site-wide WCAG 2.4.7 gap) + loading; fixed a Tailwind purge bug that dropped dynamic button classes./domains. The on-page recovery work shipped this week (entity consolidation, schema dedup, internal-linking overhaul, 16-lang re-translation). GSC on /domain-broker/* (last 30 days vs the prior 30): clicks 9 → 16, impressions 0.7K → 1.7K, and the head term "premium domain broker" went from unranked to position 19 (top-10 on its best days) — an early but real recovery curve. It also converts at 12.4%. The case against reverting is structural: the /domains/ peak predates three platform migrations + a content prune (the old site state ranked, not the path), 0 of 5,649 redirects are broken (no emergency), and a revert is a third migration that resets Google's clock for zero reclaimed equity. Recommendation: stay in place; if it stalls after 4–8 weeks, canary one low-traffic child (e.g. domain-valuation) to /domains/ — reversible — rather than flipping the whole section. (Honest caveat: single-digit-to-teens clicks and losing share as the whole site's impressions ~doubled — recovering, not fixed.)/out volume + Nord's report going forward, and stop reacting to the bowtie %./vpn/ hub post-makeover: pre-makeover baseline was 36.9% onward routing to provider pages (289/784 sessions, 7d). That's the number the redesign has to beat — first read in ~a week.