StarZero All shows Try the podcast clipper →

Dec 5, 2013 · 23m · mad

Adam Laiacano, Tumblr // Data Driven NYC 20 // Nov 2013 (Hosted by FirstMark Capital)

Adam Laiacano · 18m spoken Libby Englander · 1m spoken Matt Turck · 48s spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

At Data Driven NYC, Tumblr's Adam Laiacano presents the concept of 'data-generating products,' illustrating how intentional user interface choices naturally produce clean, high-signal datasets for personalization and recommendation systems.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 4% of the talking time here. How this is scored →

Matt as informed peer 0.2 Guest teaching 5.0 Guest disagreement 0.2 Matt pushing back 0.2
05100:0010:0020:000:53–3:37 · Matt as informed peer 0/10 Data Products vs. Data Generating Products (Elixir Case Study) Adam gives a keynote presentation defining data-generating products through the Elixir app case study. The host is not involved in this monologue segment, keeping host scores at zero.3:37–6:15 · Matt as informed peer 0/10 Concept of Tumblr's Weekly Digest Email Adam explains how digest emails from platforms like Spotify and Quora deliver high-value content based on user data. The segment is a monologue with no host participation.6:15–10:47 · Matt as informed peer 0/10 Algorithmic Logic and Personalization for Post Ranking Adam walks through Tumblr post-ranking algorithms, illustrating how normalizing engagement counts reveals true user content preferences. The host remains silent throughout the technical explanation.10:47–13:23 · Matt as informed peer 0/10 Email Customization and Tumblr's Native Classification System Adam describes how Tumblr post types function as zero-cost, high-accuracy user classification tools. Host engagement is zero during this portion of the talk.13:23–17:59 · Matt as informed peer 0/10 Other Examples of Data-Generating Features: Lists and Sentiment URLs Adam reviews other data-generating mechanisms including Foursquare lists and vanity URL sentiment tags. The monologue setup contains no host interaction.17:59–22:53 · Matt as informed peer 1/10 Concluding Remarks, Hiring Announcement, and Audience Q&A Matt Turck briefly intervenes to enforce time limits and manage a quick Q&A session. Adam politely addresses an audience member question regarding whether Tumblr post constraints homogenize user communication.0:53–3:37 · Guest teaching 5/10 Data Products vs. Data Generating Products (Elixir Case Study) Adam gives a keynote presentation defining data-generating products through the Elixir app case study. The host is not involved in this monologue segment, keeping host scores at zero.3:37–6:15 · Guest teaching 5/10 Concept of Tumblr's Weekly Digest Email Adam explains how digest emails from platforms like Spotify and Quora deliver high-value content based on user data. The segment is a monologue with no host participation.6:15–10:47 · Guest teaching 6/10 Algorithmic Logic and Personalization for Post Ranking Adam walks through Tumblr post-ranking algorithms, illustrating how normalizing engagement counts reveals true user content preferences. The host remains silent throughout the technical explanation.10:47–13:23 · Guest teaching 5/10 Email Customization and Tumblr's Native Classification System Adam describes how Tumblr post types function as zero-cost, high-accuracy user classification tools. Host engagement is zero during this portion of the talk.13:23–17:59 · Guest teaching 5/10 Other Examples of Data-Generating Features: Lists and Sentiment URLs Adam reviews other data-generating mechanisms including Foursquare lists and vanity URL sentiment tags. The monologue setup contains no host interaction.17:59–22:53 · Guest teaching 4/10 Concluding Remarks, Hiring Announcement, and Audience Q&A Matt Turck briefly intervenes to enforce time limits and manage a quick Q&A session. Adam politely addresses an audience member question regarding whether Tumblr post constraints homogenize user communication.0:53–3:37 · Guest disagreement 0/10 Data Products vs. Data Generating Products (Elixir Case Study) Adam gives a keynote presentation defining data-generating products through the Elixir app case study. The host is not involved in this monologue segment, keeping host scores at zero.3:37–6:15 · Guest disagreement 0/10 Concept of Tumblr's Weekly Digest Email Adam explains how digest emails from platforms like Spotify and Quora deliver high-value content based on user data. The segment is a monologue with no host participation.6:15–10:47 · Guest disagreement 0/10 Algorithmic Logic and Personalization for Post Ranking Adam walks through Tumblr post-ranking algorithms, illustrating how normalizing engagement counts reveals true user content preferences. The host remains silent throughout the technical explanation.10:47–13:23 · Guest disagreement 0/10 Email Customization and Tumblr's Native Classification System Adam describes how Tumblr post types function as zero-cost, high-accuracy user classification tools. Host engagement is zero during this portion of the talk.13:23–17:59 · Guest disagreement 0/10 Other Examples of Data-Generating Features: Lists and Sentiment URLs Adam reviews other data-generating mechanisms including Foursquare lists and vanity URL sentiment tags. The monologue setup contains no host interaction.17:59–22:53 · Guest disagreement 1/10 Concluding Remarks, Hiring Announcement, and Audience Q&A Matt Turck briefly intervenes to enforce time limits and manage a quick Q&A session. Adam politely addresses an audience member question regarding whether Tumblr post constraints homogenize user communication.0:53–3:37 · Matt pushing back 0/10 Data Products vs. Data Generating Products (Elixir Case Study) Adam gives a keynote presentation defining data-generating products through the Elixir app case study. The host is not involved in this monologue segment, keeping host scores at zero.3:37–6:15 · Matt pushing back 0/10 Concept of Tumblr's Weekly Digest Email Adam explains how digest emails from platforms like Spotify and Quora deliver high-value content based on user data. The segment is a monologue with no host participation.6:15–10:47 · Matt pushing back 0/10 Algorithmic Logic and Personalization for Post Ranking Adam walks through Tumblr post-ranking algorithms, illustrating how normalizing engagement counts reveals true user content preferences. The host remains silent throughout the technical explanation.10:47–13:23 · Matt pushing back 0/10 Email Customization and Tumblr's Native Classification System Adam describes how Tumblr post types function as zero-cost, high-accuracy user classification tools. Host engagement is zero during this portion of the talk.13:23–17:59 · Matt pushing back 0/10 Other Examples of Data-Generating Features: Lists and Sentiment URLs Adam reviews other data-generating mechanisms including Foursquare lists and vanity URL sentiment tags. The monologue setup contains no host interaction.17:59–22:53 · Matt pushing back 1/10 Concluding Remarks, Hiring Announcement, and Audience Q&A Matt Turck briefly intervenes to enforce time limits and manage a quick Q&A session. Adam politely addresses an audience member question regarding whether Tumblr post constraints homogenize user communication.

speaking balance: gold is Matt, purple is the guest (3 minute bins)

0:00 · Matt 4.6% · guest 95.4%0:00 · Matt 4.6% · guest 95.4%3:00 · Matt 0% · guest 100%3:00 · Matt 0% · guest 100%6:00 · Matt 0% · guest 100%6:00 · Matt 0% · guest 100%9:00 · Matt 0% · guest 100%9:00 · Matt 0% · guest 100%12:00 · Matt 0% · guest 100%12:00 · Matt 0% · guest 100%15:00 · Matt 0% · guest 100%15:00 · Matt 0% · guest 100%18:00 · Matt 3.5% · guest 96.5%18:00 · Matt 3.5% · guest 96.5%21:00 · Matt 29.1% · guest 70.9%21:00 · Matt 29.1% · guest 70.9%
Sharpest disagreement ▶ 20:55 Rebutting platform homogenization critique

Adam gently rejects the audience questioner premise that structured post formats reduce interaction complexity, arguing that boundaries instead spur unexpected creative behavior.

 Hardest push from Matt ▶ 19:08 Host enforcing schedule boundaries

Matt Turck steps in to strictly limit audience Q&A to a single question due to session overruns.

 Biggest teaching moment ▶ 8:00 Raw versus normalized engagement metrics

Adam demonstrates how raw interaction numbers misrepresent user desire, showing that data normalization reveals text and chat posts as higher preference than high-volume photos.

 Matt holds his own ▶ 19:08 Host asserting event management control

Matt Turck reasserts host authority by capping the talk and managing time constraints efficiently.
the scores for every segment, with the reasoning behind each
ChapterTopicMatt as informed peerGuest teachingGuest disagreementMatt pushing backWhy
Data Products vs. Data Generating Products (Elixir Case Study) 0500 Adam gives a keynote presentation defining data-generating products through the Elixir app case study. The host is not involved in this monologue segment, keeping host scores at zero.
Concept of Tumblr's Weekly Digest Email 0500 Adam explains how digest emails from platforms like Spotify and Quora deliver high-value content based on user data. The segment is a monologue with no host participation.
Algorithmic Logic and Personalization for Post Ranking 0600 Adam walks through Tumblr post-ranking algorithms, illustrating how normalizing engagement counts reveals true user content preferences. The host remains silent throughout the technical explanation.
Email Customization and Tumblr's Native Classification System 0500 Adam describes how Tumblr post types function as zero-cost, high-accuracy user classification tools. Host engagement is zero during this portion of the talk.
Other Examples of Data-Generating Features: Lists and Sentiment URLs 0500 Adam reviews other data-generating mechanisms including Foursquare lists and vanity URL sentiment tags. The monologue setup contains no host interaction.
Concluding Remarks, Hiring Announcement, and Audience Q&A 1411 Matt Turck briefly intervenes to enforce time limits and manage a quick Q&A session. Adam politely addresses an audience member question regarding whether Tumblr post constraints homogenize user communication.

Statements from this episode (5)

Disclosure
Tumblr recommended blogs using social graph and post content
“At Tumblr, we try to recommend blogs to people. We use things like the Social Graph. Some of the content of your posts, things like that, to find blogs that you might like.”
Adam Laiacano Dec 5, 2013 ▶ 1:07
Assertion Supported
Tumblr hosted approximately 150 million blogs as of November 2013
“There's about a hundred and fifty million of them that we host now.”
Adam Laiacano Dec 5, 2013 ▶ 5:43
Assertion Partly supported
Video, audio, and photos were Tumblr's most popular post types
“So on Tumblr you can post seven different types of content. Video, audio, photos are by far the most popular links, things like that.”
Adam Laiacano Dec 5, 2013 ▶ 7:25
Disclosure
Tumblr's email algorithm prioritized original content creators over rebloggers
“We want to give a little bit of bonus points to content creators, so if I post something and Hillary reblogs it from me, And these are both candidates to go into your email. Mine will show up instead of hers, because I came up with it.”
Adam Laiacano Dec 5, 2013 ▶ 7:47
Assertion Not checkable as stated
Tumblr's explicit post categorization yielded perfectly accurate content classification data
“If it's text, if it's a photo, whatever, so we can use this later on, and there's no predictions, there's no, oh, there's a photo in this thing, but we don't know if it's the main content of the post it's free, like, hundred percent accuracy prediction all the…”
Adam Laiacano Dec 5, 2013 ▶ 12:04
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 400 conversations transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more