What earned the audience?
Explore reach, response, and the topics connecting 192 Devoxx videos. Search a title cohort; every insight below follows your filters.
DevoxxForever · Videos tab
Case- and accent-insensitive. Use quotes for phrases and a minus to exclude a term. Ordinary terms match within words.
Where does your talk sit?
Find your talk with the title-keyword search above, then choose a matching title. Compare public reach and response, not talk quality.
Title keywords and All/Any matching narrow this alphabetical list; date, topic, and view filters do not. Choosing a talk does not change the comparison group. All videos still benchmarks against all 192.
All videos ignores title search and every other filter. Filtered mode uses the current selection, including title search.
Curve: share of comparison videos at or below each count (empirical CDF). Horizontal axis: count on a log(1 + count) scale; vertical axis: cumulative percentage. Dot: selected talk. P50/P90/P95/P99 are interpolated count thresholds, not percentile ranks.
Views and likes retain their original capture times; comments were collected separately. Comment counts can include replies and reflect YouTube's public header, not unique commenters. Different exposure, publication ages, and collection times limit comparison.
Videos leading the conversation
Top 8 within your selection. Open a title to watch the source.
The shape of attention
Video count by lifetime views. Select a band to filter the dashboard.
Reach vs. response
Do the most-watched videos also earn the highest like rates? Each dot is a video; hover or focus for details.
Like rate = likes ÷ views. Small audiences can produce unstable rates. Dot links open YouTube.
Publication cohorts
Group videos by publication date, not viewing date. Select a date to explore that cohort.
Earlier releases have had more time to accumulate views. Cohorts are not a daily view-growth series.
Topic signals
Compare audience response across title-based themes. Select a topic to filter.
Topics overlap and use title keywords only; they are not content classifications. Small topic samples are flagged.
Find your next question
Recurring words in the selected session titles, excluding apparent speaker credits and common stopwords.
Select a word to replace your title search. Counts are videos containing that exact word; the title search uses substring matching.
Video explorer
All matching videos, with exact snapshot counts. Click a column to sort. Missing comment counts sort last in either direction.
How to read this dashboard · definitions, scope & limitations
- Source: 192 publicly listed videos from DevoxxForever’s Videos tab, published from October 1, 2026 through the October 20 extraction. No qualifying uploads after October 13 were found. Shorts, Live, and unavailable uploads are excluded.
- Snapshot: exact watch-page views and public likes captured October 20, 2026, approximately 14:18–14:21 UTC. Counts are embedded locally and do not update. Individual capture times are included in CSV exports.
- Comments: collected separately from the public watch-page comments header, resolving its initial comments continuation when needed. Exact integer counts only; abbreviated/missing counts are unavailable, disabled comments are marked disabled, and request/parser failures are marked failed. None is assumed to mean zero. Each record retains comment count status, reported label, source, error (if any), and a separate UTC collection timestamp. No comment text or authors are included.
- Speaker benchmarks: all 192 videos by default, regardless of title search. Title keywords and their All/Any matching mode narrow the alphabetical talk selector, including phrases, exclusions, and case/accent normalization; other filters do not narrow it. A matching selection is preserved; otherwise the first matching talk is selected. No title matches clears the talk and its markers. Clearing the search restores all titles. Choosing a talk never changes the comparison population. You may explicitly switch to the current dashboard filters. A selected talk outside the cohort is still compared with that cohort and is marked as outside it. Comments exclude null counts and show the smaller sample size; genuine zeros are included.
- Percentiles: P50/P90/P95/P99 are count thresholds using ascending values and linear interpolation at index (n − 1) × p (the same as inclusive spreadsheet percentiles). Fractional thresholds are not observed integer counts. Percentile rank = 100 × number of comparison values ≤ the talk's value ÷ n; ties receive the same upper cumulative rank. The empirical CDF is a step curve. Empty populations yield no thresholds/ranks; a single value yields that value for every threshold. High ranks describe public counts, not talk quality.
- Publication dates: calendar dates from YouTube’s supplied publication timestamps (UTC−07:00), not event/session dates. The date interval is inclusive.
- Like rate: likes ÷ views × 100. Aggregate rates are view-weighted (sum of likes ÷ sum of views), not averages of percentages. Likes are not unique viewers or a measure of watch time.
- * Lifetime views/day: views divided by exact elapsed days between publication and each video’s capture timestamp. This is an age-normalized lifetime average, not current daily views, growth, or velocity. Older and newer videos still differ in exposure and distribution.
- Concentration: top-decile share uses the highest-viewed ceil(10% × selected count) videos. Quartiles use linear interpolation; median is the middle value (or mean of the two middle values).
- Topics: transparent, overlapping keyword rules on the title before an apparent “by” speaker credit. Expand the topic rules below to inspect them. Unmatched titles are “Other / unclassified.” Topic comparisons are descriptive, not causal.
- Limits: no impressions, click-through rate, watch time, subscriber conversions, audience demographics, traffic sources, or historical snapshots. We cannot infer content quality, changes over time, or that title wording caused performance. Dashboard metrics and insights follow your filters; speaker benchmarks use their separately selected comparison population.