CQI Referrer Attribution

1.10.0

• Note on versioning: the last publicly released version was 1.8.0. Versions 1.9.0 and 1.9.1 (same-session multi-page attribution, the is_first_touch flag, and the First Touch column) were completed as internal builds but were never publicly released. This release supersedes both and is the first public release since 1.8.0; their changelog entries below are kept as an accurate build record and are not being removed or rewritten.
• Added: Session Journey. The Sessions report now shows a Touches count for every session and, for any session with more than one page view, an expand control revealing the full page-by-page path in chronological order. Free tier, no licence gate, matching how Refer App (the standalone edition of this product) already ships this on its own Reports tab.
• Added: a live session counter on the Dashboard. The Sessions card now polls every 30 seconds and updates in place, with a brief highlight, when a new session is recorded, without a full page reload.
• Added: a “By Source” donut chart on the Dashboard, alongside the existing channel breakdown, showing session share by specific source domain rather than channel category.
• Added: a “Sessions Per Day” bar chart on the Dashboard, covering the same 30-day window as the rest of the Dashboard.
• Changed: the Dashboard’s single channel-share pie chart is now a donut, sitting alongside the new source donut and daily bar chart in a three-panel layout.
• Added: one-click sample data on the Settings tab. Loads 60 synthetic sessions spread across the past 30 days, across every channel, for a populated Dashboard, Reports, and channel share chart on a fresh install. Sample rows use a reserved session token prefix and can be removed at any time with no effect on real attribution data.

1.9.1

• Added: a “First Touch” column on the Sessions report and in its CSV export, showing whether each row is the page that first brought the visitor to your site or a page they viewed afterward in the same visit. Surfaces the is_first_touch data added in 1.9.0 directly in the free Sessions report.

1.9.0

• Added: same-session first-touch and last-touch attribution. Previously, a returning page view within the same browsing session (while the attribution cookie was still valid) was correctly attributed to the original entry channel, but no log row was written for it at all — the Sessions report only ever showed a session’s first page, never anything browsed afterward, and there was no way to see which page a same-session conversion actually happened from. Every page view in a session now gets its own row, carrying the session’s real entry channel, source, and UTM data forward, each marked with a new is_first_touch flag (true only for the session’s first page). This uses the existing daily-rotating session token exactly as it already works: no new identifier, no cross-day linking, no change to this plugin’s privacy design. Cross-day, multi-visit attribution remains explicitly out of scope.
• Fixed, audited as part of this change: every session-count query in this plugin counted rows rather than distinct sessions. With a session now able to span multiple rows, this would have inflated every session count (Dashboard, Trends, Insights’ Attribution Coverage) in proportion to how many pages each visitor browses. All four affected queries now count DISTINCT session tokens; the one query that should still count rows (the Sessions report’s own pagination total, where each row is correctly its own line) was identified and deliberately left unchanged.
• A real bug was caught and fixed before it ever shipped, by tracing the code directly rather than trusting the comment written alongside the first draft: the returning-visit code path was overwriting each new row’s landing_url with the session’s original (first-page) URL instead of the current page’s own URL, which would have made every subsequent page in a session repeat the entry page’s address forever, defeating the entire purpose of recording a row per page.
• The Plugin Test Rig gained a new scenario simulating a real two-page browsing session: it genuinely carries the attribution cookie from the first request’s response into the second request, the same way a real browser does, and confirms the second page gets its own row, the same session token, the original channel carried forward, its own distinct landing URL, and the correct is_first_touch flag on both rows.

1.8.1

• Documentation only, no code change: the “How does the AI channel work?” FAQ entry stated the default taxonomy includes 19 entries while naming only two of them, which a careful reviewer correctly flagged as potentially incomplete rather than guessing at what the rest might be. All 19 entries were checked directly against the taxonomy’s own source array: 19 domain entries cover 16 distinct platforms, since ChatGPT, Google Gemini, and Microsoft Copilot are each matched against two domains. The FAQ entry now names all 16 platforms explicitly and explains the domain-count distinction, so this never needs reverse-engineering from source again.

1.8.0

• Added: Direct / Unknown sessions are now excluded by default from the Dashboard’s Channel Share chart and summary cards, with a visible notice stating how many sessions were excluded and a one-click toggle to show them again. Direct/Unknown is usually the largest single channel and tends to visually swamp every genuinely interesting channel on a chart this size. This is a per-view, reversible default (a URL parameter, not a stored setting) — separate from the existing, permanent “excluded channels” setting on the Settings tab, which still applies underneath it and is never silently overridden by this toggle.
• Fixed during the same change: the Dashboard’s summary cards and its pie chart computed their channel breakdown independently in two separate methods, which had no mechanism keeping them in agreement. Both now apply the same exclusion logic from a single source of truth.
• Clarified, not changed: the default AI Referrer Taxonomy’s “19 entries,” stated in the 1.6.0 changelog entry below, is a count of taxonomy rows, not distinct platform names — three platforms (ChatGPT, Google Gemini, Microsoft Copilot) are each matched against two domains, so 19 entries cover 16 distinct platforms. Raised directly by a developer cross-checking the changelog against the platform names actually listed; this entry exists so that distinction is stated plainly going forward rather than left to be inferred.

1.7.0

• Added: CQI_RA_DB::get_channel_breakdown_for_period(), a public method returning a channel breakdown for an arbitrary date range. Not used by this plugin itself; added for an add-on (CQI Referrer Attribution Pro) that needs a previous-period channel breakdown to show attribution coverage change over time, which get_dashboard_metrics() does not provide on its own.