StarZero All shows Try the podcast clipper →

Oct 21, 2015 · 25m · mad

Liz Crawford, Birchbox // Data Science & Analytics at Birchbox (Hosted by FirstMark Capital)

Liz Crawford · 19m spoken Matt Turck · 1m 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 a DataDrivenNYC event, Birchbox CTO Liz Crawford details how the beauty subscription retailer builds its data organization, deploys explicit and implicit personalization algorithms, and designs cross-channel event tracking systems. Crawford also shares actionable takeaways on building data capabilities incrementally and integrating data scientists directly into product development teams.

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 5.2% of the talking time here. How this is scored →

Matt as informed peer 0.8 Guest teaching 0.8 Guest disagreement 0.2 Matt pushing back 0.4
05100:0010:0020:000:51–5:14 · Matt as informed peer 0/10 Evolution of Birchbox's Data Organization Structure This segment is a presentation monologue by Liz Crawford explaining Birchbox's organizational structure and definitions of data science roles. Because the host does not participate, host-side metrics are zero.5:14–8:20 · Matt as informed peer 0/10 Explicit Personalization: The Subscription Box Optimization Problem Liz continues her solo presentation detailing the algorithmic optimization of monthly subscription box allocation. No host interaction occurs during this segment.8:20–10:50 · Matt as informed peer 0/10 Online Recommendation Engines and Personalized Storefronts Liz presents on Birchbox's storefront personalization and event tracking infrastructure. As a pure monologue segment, host scores remain zero.10:50–14:28 · Matt as informed peer 0/10 Attentive Messaging and Triggered CRM Communications Liz wraps up her talk with insights on triggered CRM messaging and key organizational takeaways. The segment is entirely monologue.14:28–25:41 · Matt as informed peer 4/10 Q&A Session with Matt Turck and Audience Discussion Matt Turck opens the Q&A with informed questions about early-stage startup hiring and cultural integration, while audience members ask technical questions. Liz answers collaboratively, offering educational insights on hiring scrappy data scientists and build-versus-buy trade-offs.0:51–5:14 · Guest teaching 0/10 Evolution of Birchbox's Data Organization Structure This segment is a presentation monologue by Liz Crawford explaining Birchbox's organizational structure and definitions of data science roles. Because the host does not participate, host-side metrics are zero.5:14–8:20 · Guest teaching 0/10 Explicit Personalization: The Subscription Box Optimization Problem Liz continues her solo presentation detailing the algorithmic optimization of monthly subscription box allocation. No host interaction occurs during this segment.8:20–10:50 · Guest teaching 0/10 Online Recommendation Engines and Personalized Storefronts Liz presents on Birchbox's storefront personalization and event tracking infrastructure. As a pure monologue segment, host scores remain zero.10:50–14:28 · Guest teaching 0/10 Attentive Messaging and Triggered CRM Communications Liz wraps up her talk with insights on triggered CRM messaging and key organizational takeaways. The segment is entirely monologue.14:28–25:41 · Guest teaching 4/10 Q&A Session with Matt Turck and Audience Discussion Matt Turck opens the Q&A with informed questions about early-stage startup hiring and cultural integration, while audience members ask technical questions. Liz answers collaboratively, offering educational insights on hiring scrappy data scientists and build-versus-buy trade-offs.0:51–5:14 · Guest disagreement 0/10 Evolution of Birchbox's Data Organization Structure This segment is a presentation monologue by Liz Crawford explaining Birchbox's organizational structure and definitions of data science roles. Because the host does not participate, host-side metrics are zero.5:14–8:20 · Guest disagreement 0/10 Explicit Personalization: The Subscription Box Optimization Problem Liz continues her solo presentation detailing the algorithmic optimization of monthly subscription box allocation. No host interaction occurs during this segment.8:20–10:50 · Guest disagreement 0/10 Online Recommendation Engines and Personalized Storefronts Liz presents on Birchbox's storefront personalization and event tracking infrastructure. As a pure monologue segment, host scores remain zero.10:50–14:28 · Guest disagreement 0/10 Attentive Messaging and Triggered CRM Communications Liz wraps up her talk with insights on triggered CRM messaging and key organizational takeaways. The segment is entirely monologue.14:28–25:41 · Guest disagreement 1/10 Q&A Session with Matt Turck and Audience Discussion Matt Turck opens the Q&A with informed questions about early-stage startup hiring and cultural integration, while audience members ask technical questions. Liz answers collaboratively, offering educational insights on hiring scrappy data scientists and build-versus-buy trade-offs.0:51–5:14 · Matt pushing back 0/10 Evolution of Birchbox's Data Organization Structure This segment is a presentation monologue by Liz Crawford explaining Birchbox's organizational structure and definitions of data science roles. Because the host does not participate, host-side metrics are zero.5:14–8:20 · Matt pushing back 0/10 Explicit Personalization: The Subscription Box Optimization Problem Liz continues her solo presentation detailing the algorithmic optimization of monthly subscription box allocation. No host interaction occurs during this segment.8:20–10:50 · Matt pushing back 0/10 Online Recommendation Engines and Personalized Storefronts Liz presents on Birchbox's storefront personalization and event tracking infrastructure. As a pure monologue segment, host scores remain zero.10:50–14:28 · Matt pushing back 0/10 Attentive Messaging and Triggered CRM Communications Liz wraps up her talk with insights on triggered CRM messaging and key organizational takeaways. The segment is entirely monologue.14:28–25:41 · Matt pushing back 2/10 Q&A Session with Matt Turck and Audience Discussion Matt Turck opens the Q&A with informed questions about early-stage startup hiring and cultural integration, while audience members ask technical questions. Liz answers collaboratively, offering educational insights on hiring scrappy data scientists and build-versus-buy trade-offs.

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

0:00 · Matt 0% · guest 100%0:00 · Matt 0% · guest 100%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 16.9% · guest 83.1%12:00 · Matt 16.9% · guest 83.1%15:00 · Matt 17.8% · guest 82.2%15:00 · Matt 17.8% · guest 82.2%18:00 · Matt 7.3% · guest 92.7%18:00 · Matt 7.3% · guest 92.7%21:00 · Matt 1.2% · guest 98.8%21:00 · Matt 1.2% · guest 98.8%24:00 · Matt 2.3% · guest 97.7%24:00 · Matt 2.3% · guest 97.7%
Sharpest disagreement ▶ 19:40 Liz politely pushing back on audience premise

Liz directly but politely addresses a confused audience question regarding statistical confidence by noting that she might be missing the point after explaining their approach to A/B testing and model impact.

 Hardest push from Matt ▶ 14:29 Matt probing the chicken-and-egg startup dilemma

Matt pushes on Liz's takeaway by framing the classic VC dilemma of whether startups should hire data scientists before or after possessing data.

 Biggest teaching moment ▶ 14:58 Liz explaining early data science realities

Liz clarifies that early data scientists at startups need to write production code and be scrappy enough to instrument and gather data themselves rather than waiting for existing datasets.

 Matt holds his own ▶ 15:51 Matt highlighting the ivory tower organizational trap

Matt demonstrates industry insight by pointing out the widespread friction startups face when data scientists operate in isolated ivory towers away from engineering teams.
the scores for every segment, with the reasoning behind each
ChapterTopicMatt as informed peerGuest teachingGuest disagreementMatt pushing backWhy
Evolution of Birchbox's Data Organization Structure 0000 This segment is a presentation monologue by Liz Crawford explaining Birchbox's organizational structure and definitions of data science roles. Because the host does not participate, host-side metrics are zero.
Explicit Personalization: The Subscription Box Optimization Problem 0000 Liz continues her solo presentation detailing the algorithmic optimization of monthly subscription box allocation. No host interaction occurs during this segment.
Online Recommendation Engines and Personalized Storefronts 0000 Liz presents on Birchbox's storefront personalization and event tracking infrastructure. As a pure monologue segment, host scores remain zero.
Attentive Messaging and Triggered CRM Communications 0000 Liz wraps up her talk with insights on triggered CRM messaging and key organizational takeaways. The segment is entirely monologue.
Q&A Session with Matt Turck and Audience Discussion 4412 Matt Turck opens the Q&A with informed questions about early-stage startup hiring and cultural integration, while audience members ask technical questions. Liz answers collaboratively, offering educational insights on hiring scrappy data scientists and build-versus-buy trade-offs.

Statements from this episode (10)

Insight
Centralized functional data teams prevent duplicate work and foster peer learning
“To organize functionally, but what we've found is that people can learn from each other really well that the data scientists and the statistical analysts can learn from each other and that that's great, and that it really helps when people are working together…”
Liz Crawford Oct 21, 2015 ▶ 1:45
Disclosure
Birchbox defines data scientists as PhD-level coders on product teams
“So we see a data scientist as someone who contributes to product development, essentially. Someone with very specialized skills, Who helps us deliver our products into market. And so by specialized skills, I essentially mean a PhD or something very much like i…”
Liz Crawford Oct 21, 2015 ▶ 2:26
Disclosure
Birchbox never sends subscribers the same sample product twice
“We never send you the same thing twice, for example.”
Liz Crawford Oct 21, 2015 ▶ 6:07
Disclosure
Birchbox uses the Gurobi solver for monthly subscription box personalization
“And we do this with the help of a very nice solver called Garobi.”
Liz Crawford Oct 21, 2015 ▶ 7:04
Assertion Not checkable as stated
Sample selection is Birchbox's highest monthly customer engagement moment
“So every month our customers are able to choose one of their samples. This is our highest engagement moment every month.”
Liz Crawford Oct 21, 2015 ▶ 7:47
Assertion Not checkable as stated
Birchbox saw huge conversion gains by shifting to triggered messaging
“And moving more and more and more and more of the messages we send our customers over push, over email to triggered, we've seen huge increases in the conversion of those messages.”
Liz Crawford Oct 21, 2015 ▶ 11:52
Insight
Data science teams must be empowered to gather new data
“Don't be limited by the data that you have today, and if you run a data science team, don't let them be limited by it. Empower people to go out and get the data that they need to do their jobs.”
Liz Crawford Oct 21, 2015 ▶ 14:01
Insight
Non-scrappy data scientists do not belong in early-stage startups
“I wouldn't hire someone that couldn't go out and be scrappy and get their hands dirty. That person doesn't belong in an early stage startup anyway.”
Liz Crawford Oct 21, 2015 ▶ 15:42
Assertion Not checkable as stated
Birchbox's catalog recommendation algorithm boosted full-size product conversions by 20%
“We were using one of our recommendation algorithms just recently in a catalog sort which we newly put in, and it showed a 20% lift in conversion on, like, full size purchase.”
Liz Crawford Oct 21, 2015 ▶ 20:22
Assertion Not checkable as stated
Birchbox trained nearly its entire company workforce to write SQL
“Trained almost the entire company to use SQL.”
Liz Crawford Oct 21, 2015 ▶ 24:22
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