Landing page change log

This log connects landing-page iterations with the events in events.duckdb. Localhost and 127.0.0.1 sessions are excluded from visitor baselines.

2026-07-26 — Shorter, article-first landing page

Baseline

Data available through 2026-07-26 (the latest /landing.html event is from 2026-07-25). The current Teenage Engineering design launched on 2026-07-13.

40 non-local sessions since launch.
Acquisition: 20 Twitter/X, 12 direct, 8 internal.
45% reached the footer (18 of 40).
Median maximum scroll was 65.5%.
Median observed duration was 16 seconds.
25 sessions generated a click event, but most article clicks could not be attributed: the tracker only captured an href when the exact clicked element was the anchor. Clicking nested article titles produced generic EM, DIV, or P events.
The page exposed 39 latest-article rows after the profile and two newsletters.

Hypothesis

Visitors—half of whom arrive from Twitter/X—should understand who Javi is in the first viewport and encounter a small, useful set of articles before newsletter promotion. Reducing choice and page length should increase attributable article visits and section completion.

Changes

Rewrote the hero around a direct proposition: co-founder of Tinybird, data-product builder, and writer about engineering, product, and startups.
Added three prominent “start here” articles selected from representative posts with strong traffic in the previous 90 days.
Moved writing ahead of the about and newsletter sections.
Reduced the dynamic latest list from 39 to 9 articles.
Removed duplicated profile/spec content and replaced it with a shorter career narrative.
Made newsletter language and subject matter explicit.
Added a responsive landing-specific layout while preserving the existing visual system.
Added a page meta description and normalized external links to HTTPS.
Added named analytics sections and stable data-analytics labels.
Fixed click attribution to record the nearest anchor, even when a nested element is clicked.
Fixed isNewMax, which was always recorded as false because the maximum was updated before the comparison.

Metrics to compare after deployment

Use only sessions after the deployment timestamp and exclude local referrers.

Article click-through rate by label, especially featured-* and latest-*.
Section entry rate for start-here, latest-writing, about, and newsletter.
Newsletter click-through rate by language.
Median duration, maximum scroll, and footer completion.
Performance split by source and viewport width.

2026-07-27 — Reliable, versioned landing measurement

Deployment timestamp: pending. This iteration is not deployed yet.

Baseline

Data window: 2026-07-26 08:50 CEST through 2026-07-27 09:18 CEST, beginning with the first event after commit 5cba1e4. Localhost and 127.0.0.1 referrers are excluded.

4 sessions, all direct: 3 desktop and 1 mobile. This is too small a sample for content or layout conclusions.
Median observed duration was 84 seconds among the 3 sessions with an end event.
Median maximum scroll was 50%; 2 of 4 sessions reached the end.
Named section entry was recorded for 1 of 4 sessions for each of start-here, latest-writing, about, newsletter, and footer.
There were no attributable article, newsletter, or social clicks. One internal “about” jump was attributable by destination, but its stable label was missing.
The three featured articles remain the strongest relevant choices in the 90-day window from 2026-04-28 through 2026-07-27: 63 sessions for the Google-front-page article, 35 for the four-years-of-data-engineering article, and 17 for “40 things I learned about data.”
Measurement quality is not yet suitable for comparing click-through rates. The deployed HTML has stable labels, but a browser served the pre-fix tracker from cache: the recorded anchor click had an href but no label. Section visibility also required more than half of a section to be on screen, which undercounts sections taller than the viewport.

Hypothesis

With the post-redesign sample too small for another visual change, preserving the current article-first page and making its measurement reliable will produce a valid comparison sooner than changing content again. Versioned tracker payloads, cache-busted delivery, and viewport-relative section thresholds should make the next session cohort distinguishable and its important interactions attributable.

Changes

Added analytics version 2026-07-27 to every tracker payload so this cohort can be queried independently.
Cache-busted the tracker URL to ensure visitors receive nearest-anchor click attribution and stable data-analytics labels.
Changed section entry to require a meaningful 25% overlap capped at 25% of the viewport, allowing tall sections to register without treating a one-pixel intersection as entry.
Preserved selected and copied excerpts, capped at 100 characters, as an editorial signal alongside the full selection length.
Added landing-scoped visible keyboard focus and disabled the landing-row transition when the visitor prefers reduced motion.
Kept landing content, featured articles, and section order unchanged because the 4-session sample does not justify another redesign.

Metrics to compare after deployment

Use sessions with analyticsVersion = '2026-07-27', starting at the confirmed deployment timestamp, and keep excluding local referrers.

Coverage of analyticsVersion, label, href, and sectionName.
Article click-through rate for featured-* and latest-*.
Section entry rate for start-here, latest-writing, about, newsletter, and footer.
Newsletter and social click-through rates.
Median duration, maximum scroll, and reached-end rate, split by source and viewport.
Review selected and copied excerpts only in aggregate to identify repeatedly useful passages.

2026-07-28 — Clean landing experiment boundary

Deployment timestamp: pending. This iteration is not deployed yet.

Baseline

Data window: 2026-07-27 10:36 CEST through 2026-07-28 18:34 CEST, covering the 15 non-local-referrer sessions carrying analyticsVersion = '2026-07-27'. The window begins with the first observed versioned event; it is not a confirmed deployment timestamp. Repeated batched events were deduplicated by session and event payload.

Acquisition: 7 direct, 4 internal, 3 Twitter/X, and 1 search session.
13 sessions were desktop and 2 were narrow/mobile, too few for a meaningful device comparison.
Median observed duration was 18 seconds among the 11 sessions with an end event. Median maximum scroll was 18%, and 3 of 15 sessions reached the footer/end.
Section entry was 14 of 15 for start-here, 8 of 15 for latest-writing, 5 of 15 for about, 4 of 15 for newsletter, and 3 of 15 for footer.
Five sessions opened an article: 3 from the featured set and 2 from latest writing. There were no newsletter or social clicks. Six of six link clicks carried both a stable label and a destination; one additional click event was a non-link interaction.
In the 90-day window from 2026-04-29 through 2026-07-28, the current featured articles remained the strongest relevant choices: 38 sessions for the four-years-of-data-engineering article, 23 for the Google-front-page article, and 20 for “40 things I learned about data.”
The database contains clearly local filesystem paths. Direct local visits have no localhost referrer, so referrer filtering cannot remove all development traffic from the current cohort.

Hypothesis

The current content is already producing article visits, and 15 sessions are not enough evidence for another layout or copy change. Preventing local collection, recording only actionable link clicks, and emitting one reliable end event will create a clean post-change cohort without disrupting the article-first experience.

Changes

Kept the landing content, order, and responsive layout unchanged.
Added analytics version 2026-07-28 and cache-busted the tracker URL.
Stopped analytics on file:, localhost, IPv4 loopback, and IPv6 loopback pages.
Limited click events to links, preserving nearest-anchor destination and stable label attribution.
Clamped scroll progress to 0–100% and reports 100% for pages without a scrollable area.
Moved session-end capture to pagehide and guarded it so one page view emits at most one end event.
Stopped sending selected or copied text; only aggregate-safe character counts are retained.

Metrics to compare after deployment

Use sessions with analyticsVersion = '2026-07-28', beginning at the confirmed deployment timestamp.

Confirm that local and filesystem paths no longer appear in the versioned cohort.
Coverage of label and href across all click events.
Coverage and uniqueness of end events, plus median duration and maximum scroll.
Article click-through rate for featured and latest writing.
Section entry and footer completion rates.
Newsletter and social click-through rates.

2026-07-31 — Direct article CTA and denser first viewport

Deployment timestamp: 2026-07-31 18:48 CEST (commit b4576fe).

Baseline

Data window: versioned non-local sessions with analyticsVersion in ('2026-07-27', '2026-07-28') through 2026-07-31 14:32 UTC (latest event in events.duckdb). Localhost and 127.0.0.1 referrers are excluded. Batched events were expanded and deduplicated by session, event fields, and second.

26 sessions: 14 direct, 6 Twitter/X, 4 internal, 2 search. Only 2 mobile viewports (viewportWidth < 700); treat device splits as unreliable.
Median observed duration 18s; median maximum scroll 18%. 6 of 26 reached the footer/end. 11 of 26 recorded 0% max scroll; 13 of 26 ended in ≤5s.
Section entry (versioned): start-here 23/26 (88%), latest-writing 13/26 (50%), about 7/26 (27%), newsletter 6/26 (23%), footer 5/26 (19%).
7 of 26 sessions had a click. Attributable article labels included featured-four-years (strongest), latest-learn-to-cook, featured-forty-things, and featured-google-frontpage. No newsletter or social clicks in this window.
Measurement quality for 2026-07-28: no file: paths; 18 of 19 click events carried both label and href. End events still include long backgrounded tabs (multi-hour duration); prefer medians.
90-day article traffic (from 2026-05-01, non-local, content paths): featured set remains appropriate—four-years (37 sessions), 40-things (20), Google frontpage (18). Popular non-featured pieces include inspiration notes (experience-in-movies, learn-to-cook) and sql-agent; kept out of “start here” to preserve a clear professional introduction.

Hypothesis

Most visitors bounce before scrolling past the first block. The primary hero CTA only jumped to #start-here, which is often already on screen, so it could not create an article visit. Putting a direct link to the strongest featured article in the first viewport, tightening hero density so writing stays visible, and shortening the latest list from 9 to 5 should raise article CTR among short sessions without changing who Javi is or what the page is for.

Changes

Primary hero CTA is now a direct link to the four-years data-engineering article (hero-read-four-years), styled as a solid button.
Secondary hero links: in-page “more writing” and “about me” (no longer the primary action).
Slightly shorter hero copy; reduced hero type size and vertical spacing so featured cards sit higher in the first viewport.
Marked the top featured card as recommended (te-featured-primary) and shortened card blurbs; reduced featured min-height.
Cut the dynamic latest list from 9 to 5 articles.
Analytics version 2026-07-31 with cache-busted tracker URL for a clean post-change cohort.

Metrics to compare after deployment

Use sessions with analyticsVersion = '2026-07-31', beginning at the confirmed deployment timestamp; exclude local referrers.

Article CTR overall and for hero-read-four-years vs featured-* vs latest-*.
Share of sessions with any article click among those with duration ≤15s.
Section entry for start-here and latest-writing; footer completion.
Median duration and maximum scroll (ignore extreme open-tab durations).
Newsletter and social CTR (still secondary).

2026-08-03 — Name-first hero and active duration

Deployment timestamp: 2026-08-03 17:31 CEST (this commit).

Baseline

Data window: non-local sessions with analyticsVersion = '2026-07-31' from 2026-07-31 18:51 CEST through 2026-08-03 14:53 CEST (latest event in events.duckdb at analysis time). Localhost and 127.0.0.1 referrers excluded. Batched events expanded and deduplicated by session, fields, and second. Wall-clock durations above 1 hour treated as open-tab outliers for medians.

18 sessions: 13 Twitter/X (t.co), 2 direct, 2 internal, 1 empty referrer. Zero mobile viewports (viewportWidth < 700); device splits still unreliable.
Relative to the pre-change versioned cohort (2026-07-27 + 2026-07-28, n=38): article session CTR 22.2% vs 13.2%; featured CTR 16.7% vs 7.9%; hero article CTA 11.1% (new); newsletter CTR 11.1% vs 0%; section entry much deeper (latest-writing 72% vs 37%, about 78% vs 18%, footer 61% vs 13%). Mix shifted toward Twitter, so treat lift as directional, not proven.
Median duration among end events under 1h: 19s. Median max scroll 100% (end-event maxProgress or scroll progress). 13/18 reached footer/end.
Short sessions (duration ≤15s): 3 of 5 had an article click (vs 1 of 21 pre).
Top landing click labels by sessions: hero-about (3), featured-forty-things (3), featured-four-years (2), featured-google-frontpage (2), hero-read-four-years (2). First newsletter and social clicks appeared in this cohort.
Measurement quality: labels and hrefs present on link clicks. Wall-clock duration still polluted by multi-hour backgrounded tabs (e.g. >20k seconds). Some reached_end / 100% scroll pairs with near-zero end maxProgress remain hard to interpret.
90-day content traffic (from 2026-05-05, non-local): featured set still appropriate for a professional intro—four-years (32 sessions), 40-things (21), Google frontpage (14). Strong non-featured pieces (learn-to-cook, experience-in-movies, sql-agent) stay out of “start here.”

Hypothesis

The direct article CTA is working, especially on short Twitter sessions, but the most-clicked control is still hero-about: visitors arrive needing a clear name and role before they commit to a post. Putting the name in the H1 and a compact who/what line in the first viewport—while keeping the article CTA and making its label name the destination—should raise article CTR without another layout rewrite. Recording visibility-aware activeDuration will make the next cohort’s engagement medians trustworthy.

Changes

Hero H1 is now the name (“Javi Santana”); location stays in the kicker; lede states co-founder role plus what he builds and writes about.
Primary CTA copy names the destination article (read: four years of data engineering →); label hero-read-four-years unchanged.
Slightly tighter hero spacing and name type scale so the first viewport stays article-forward on desktop and mobile.
On narrow viewports, stack the primary CTA above secondary links and allow the button label to wrap so it does not clip at 320px width.
Page title/description lead with the name for share/tab context.
Featured set, latest list length, section order, and visual system unchanged.
Analytics version 2026-08-03 with cache-busted tracker URL.
End events include activeDuration (seconds the document was visible); reached_end prefers [data-analytics-section="footer"].

Metrics to compare after deployment

Use sessions with analyticsVersion = '2026-08-03', beginning at the confirmed deployment timestamp; exclude local referrers. Prefer activeDuration for engagement medians; keep wall-clock duration only for outlier checks.

Article CTR overall and for hero-read-four-years vs featured-* vs latest-*.
Rate of hero-about clicks (should fall if the hero answers identity).
Share of sessions with any article click among those with activeDuration ≤15s.
Section entry for start-here, latest-writing, about; footer completion.
Newsletter and social CTR (secondary).
Coverage of activeDuration on end events.