Palette Data Analyzer

Changelog

Older releases are listed in changelog.txt inside the plugin folder.

4.20.4

• Added the dedicated plugin page as the Plugin URI.

4.20.3

• Removed the redundant Plugin URI header so the plugin and author URIs are not identical.

4.20.2

• Clarified which capabilities belong to the free plugin and which require the separately distributed Pro add-on.
• Connection and relay tests now run only after an administrator explicitly clicks a nonce-protected test link; opening Settings never makes a network request.
• Removed Pro upgrade cards from the free plugin dashboard; the single contextual add-on notice remains on the Settings page.

4.20.1

• Google Search Status updates are now strictly opt-in. The upload screen explains the network request and the feed does not include the site URL in its User-Agent.
• Removed code that suppressed other plugins’ admin notices.
• Added the bundled Chart.js MIT license and moved the WordPress minimum version into the main plugin header.

4.20.0

• The Action Plan is now ordered by estimated impact inside each priority level, instead of by the order the modules happen to run in. Previously every item of one module came out in a block, so a report with ten page-two queries opened with ten near-identical lines while a single low-CTR page worth several times the traffic sat further down. Each action is now scored in one common unit – the clicks it would gain or recover, using your own site’s CTR curve for the target position – and ties are broken by title so the same file still produces the same report twice.

4.19.0

• A query or page that did not exist in the previous period no longer shows a position drop. With no earlier rank, its “position change” was computed as zero minus the current rank, so a brand-new query at position 15.3 was reported as having fallen 15.30 places. Those rows are now labelled “new”.
• The dashboard said “19 analysis modules” in both editions. It now states the real count for the edition you are running.

4.18.0

• Search Console reports the same page once per anchor link, so /page/, /page/#h-1 and /page/#h-2 arrived as three separate rows. On a real site 376 of 585 page rows were anchors and a single page was split across eleven of them, which meant no row carried that page’s real weight and one page could fill most of the action plan. Anchors are now merged: clicks and impressions are summed, position becomes an impression-weighted average, and CTR is recalculated from the totals rather than averaged.

4.17.0

• Saving a report now returns its archive ID, so anything referring back to that exact report — such as a shareable link — can name the right row.
• Setup guide updated for Google’s new console: “Publishing status” and “Test users” both moved to the Audience page, and the wizard links there directly.

4.16.0

• Fixed a slowdown on servers that cannot reach Google. The Google Search Status feed is fetched while building a report, but a failed fetch was never cached — so every single report waited out the full 15-second timeout, again and again. Failures are now remembered for an hour, the timeout is 5 seconds, and the previous data is kept rather than discarded.
• That feed can now go through the relay. It was missing from the allowlist, so a configured relay did not help it. Added to both the PHP list and the sample Cloudflare Worker.

4.15.0

• The setup guide now covers the “Access blocked / Error 403: access_denied” screen people hit when connecting their Google account, and adds a step for publishing the OAuth app. That step matters more than it looks: while the app stays in Testing, Google expires the connection after seven days, so the plugin silently loses access about once a week and scheduled reports stop. Publishing removes that limit and needs no Google review.

4.14.1

• Bot setup instructions no longer print bare @handles, which some marketplace scanners flag as personal contact details. The instructions name the official BotFather bot instead.

4.14.0

• The settings page no longer probes Google on load. It only checks connectivity once you have entered a Client ID or a property address — before that there is nothing to connect to, and probing meant an outbound request just for opening a page.
• Sections that need a direct API connection now say so instead of showing a “Pro” upsell. Those sections have no data when the report comes from a file upload, and telling a customer who already owns Pro to buy Pro was confusing.
• Added a PALDA_DISTRIBUTION constant so a vendor-neutral build can be produced from the same source, without maintaining a second copy of the plugin.

4.13.0

• A fresh install no longer makes any outbound request until you generate your first report. The Google Search Status feed was fetched on a daily cron that scheduled itself five minutes after the first admin page view — before the user had done anything, and while its output had nothing to label yet. It is now scheduled only once a report exists. Existing installs keep their schedule.

4.12.0

• Fixed “jump to section”, which had been broken in Persian. Section IDs were built from the section number — a translated string. In English it stayed “1” and worked; in Persian it became “۱”, every non-ASCII character was stripped, and all twenty sections ended up with the identical id pda-sec-. So getElementById always returned the first section (every jump went to the top of the report) and the active-section highlight had nothing to map to. IDs are now derived from section order and are language-independent.
• Pro teaser sections had the same bug and were also missing from the jump bar entirely, so free users could not jump to them. Both section builders now share one ID source.
• The jump offset was hardcoded at 120px while the real sticky-bar height is around 230px, so the target landed hidden underneath the bar. It is now measured, which also covers RTL, mobile and a hidden admin bar.
• Jumping now survives environments that ignore smooth scrolling (reduced-motion settings), re-checks its position once the scroll settles — charts change height as they load — and cancels a pending correction when you click another section.
• Bidirectional text is now derived from content, not assumed. Every link was forced to direction: ltr, which is right for a URL but wrong for links whose text is Persian, whose trailing punctuation jumped to the wrong side. Links and inline code now use unicode-bidi: plaintext, so each element takes its direction from its own first strong character. Standalone URLs stay explicitly LTR so a Persian slug cannot flip them.
• Multi-line URLs in tables were right-aligned line by line, breaking their left-to-right reading order. They are block-level now, so the box follows the page direction while the text inside stays left-aligned.
• Free dashboard: the “Get started” link pointed at the settings page, which without a licence is only a Pro pitch — a free user clicking it landed on a sales page instead of somewhere they could work. It now points at the file upload, and the main button carries a “Start here” badge until the first report exists.

4.11.0

• Plugin Check compliance: zero errors and zero warnings against the official WordPress plugin-check tool.
• readme.txt is now in English, as the plugin directory requires. The full Persian history moved to changelog.txt.
• File operations in the uninstaller and the ZIP-signature check now use WP_Filesystem, so they work on hosts where WordPress writes over FTP or SSH.
• Dropped wp_set_option_autoload(), which needs WordPress 6.4 while this plugin supports 5.8. The same result is achieved with functions available in every supported version.
• Removed the manual load_plugin_textdomain() call; WordPress has loaded translations automatically since 4.6. Verified that Persian still loads without it.

4.10.0

• The palda_settings_sections seam is now live, so the Pro add-on can render its own settings block without any of its code living in the free plugin.
• PALDA_Connectivity::rewrite_google() renamed to rewrite_if_allowed(); the relay allowlist is no longer Google-only. The old name is kept as an alias.
• Sample Cloudflare Worker: api.telegram.org added to the allowlist.

4.9.0

• Report thresholds now scale with the length of the exported window. Minimum-impression gates were tuned for 28 days, so on a 3-month export they were roughly three times too loose and on a 7-day export three times too strict.

4.8.0

• Comparison exports now work with any date range, not just 28 days. Previously the column detection matched the literal text “Last 28 days”; every other range silently produced a report where all numbers were zero, alongside a misleading “your file is single-period” message.
• Added a guard against that silent failure: if a sheet has rows but no numbers could be read, the plugin says so and names the sheet.
• Re-running the same report no longer clutters the archive. Each run used to insert a new row, which put duplicate points into the time series and dragged volatility toward zero.

4.7.0

• Rewrote the ranking-volatility module. It previously measured the absolute difference between two periods, which is not volatility and duplicated the “biggest movers” module. It now uses the standard deviation of position across at least three archived reports.
• Each report now stores a snapshot of the top 50 pages so per-page time series can be built.

4.6.1

• Property selection in the report archive is now correctly limited to the Pro add-on.
• Fixed archive rows saved under an inconsistent property format, with a one-time migration.