Brothers in Bayes: Moving things forward is the only thing that matters

Written by Finn Höner, PhD Candidate in Quantitative Marketing at the Erasmus School of Economics

The birth of a scientific collaboration

It was 2007, and Joachim Büschken had invited Thomas Otter to give a seminar on Bayesian statistics at his university, the Katholische Universität Eichstätt-Ingolstadt, Germany. Otter posed a condition: Otter would come if he could bring Greg Allenby, a leading figure in Bayesian marketing models, along. For Büschken, this wasn't a problem, but an honor. After the seminar came dinner at Büschken's home, where both the conversation and the wine flowed freely. That dinner marked the beginning of a major research partnership. Since then, Allenby and Büschken  have published many influential papers in top journals, including four on topic models for text analytics, with more still in progress.

Their latest paper “A topic change point model for phrase-based topic inference”, recently published in IJRM, introduces a new type of topic model for text analytics with several innovations. Topic models are the bread and butter of any marketing researcher interested in analyzing social media comments, product reviews, or customer service interactions at a large scale, with the purpose of identifying latent drivers behind these texts. The proposed topic model looks at word sequences to find where topics change in text and uses grammar to help identify these boundaries. This continues their research on text analytics, which began when they "stumbled onto" Latent Dirichlet Allocation (LDA) during Joachim's first visit to The Ohio State University.

From a paper to a research stream

The collaboration didn't begin with topic models. After that first dinner, Allenby invited Büschken to Columbus for Ohio State's PhD seminar on Bayesian statistics, where they discovered a shared interest in analyzing customer reviews. Both were curious, but neither knew quite how to approach this analysis. After thinking a lot about analyzing texts and exploring existing approaches from different disciplines, they arrived at Latent Dirichlet Allocation as the method to use.

Their first published paper went through extensive revisions during review and required them to build and implement three different bespoke LDA models over the course of the review process. This means three times of writing all the code, testing everything, and rewriting the manuscript. The process was gruelling, but successful. "From there," Büschken notes, "the next paper in the sequence was just obvious." The research stream had begun, and the flow shows no signs of stopping.

Complementary forces in action

What makes this collaboration work? Both emphasize complementary strengths. Together they can "really zero in on the line of thought that's the contribution". Allenby sees Büschken as "super curious and humble," a driving force behind their collaborative energy. Their process, Allenby explains, is like being "sharks in the water, swimming there with the data and trying out different stuff." They explore until they "catch a smell of what the paper is supposed to look like, and then we're really tracking that smell down."

Through this process, they've learned an important lesson: "Coding is not the bottleneck, positioning is," Büschken reflects. Getting the positioning of a paper right—that's the real intellectual challenge, as positioning their contributions within the broader theoretical and practical landscape requires something beyond technical skill.

“Coding is not the bottleneck, positioning is.” - Greg Allenby & Joachim Büschken

A real relationship, not a transaction

Ask either of them what makes their partnership special, and you'll hear the same phrase: "it is a real relationship." Both emphasize that they put no expectations on each other and are simply grateful to be able to research together. Allenby puts it in spiritual terms: "Neither one of us owns the paper or owns the idea. We're just participating in it." They call each other "brothers in Bayes," viewing their extensive research output not as the goal, but as "a byproduct of our human relationship."

This selfless perspective extends to handling challenges. Allenby mentions an upcoming visit from Büschken to Ohio State when Allenby will be absent, although he originally extended the invitation. There are no hard feelings, no strings attached. "Joachim is welcome to join, and it's just a nice thing that he's here", Allenby says simply. They have, as both note, a respectful relationship, but do not flatter one another's ego. There's friendly bantering, direct feedback, and a shared commitment to moving forward.

"How do we move it forward? Everything else doesn't matter." - Greg Allenby & Joachim Büschken

"How do we move it forward? Everything else doesn't matter." That's the mantra they work by, both point out. When relationships become transactional, that's when they sour. "Transactional love kills," Allenby observes. They've both experienced this earlier in their careers and can detect transactional behavior quickly. In their collaboration, they've avoided this trap entirely.

"We don't own the idea," they explain. "It's really about moving [the science] forward." They feel privileged to do the research they do, and publishing is "the icing on the cake," not the goal itself.

Looking ahead

Both are excited about current research closely tied to volunteer work Allenby does to help inmates transition out of prison. A new paper employing another version of topic modelling analyzes parole hearing transcripts.  It's high-stakes research—decisions about people's freedom—which both find deeply meaningful. This project went from a blank sheet of paper to acceptance within months.

When asked how others might build similar collaborations, both demur. Büschken admits: "If you asked me to build a collaboration like that with somebody else, I wouldn't know how." But there's something more than luck here, a shared commitment to humility, generosity, and genuine curiosity that has sustained this partnership across continents and decades.

With new papers already in progress, both remain focused on the same thing: moving the research forward, together. The papers are remarkable, but their human relationship makes it all possible.

---

Meet Joachim Büschken and Greg Allenby

If you were not a marketing researcher, what would you be?

Joachim Büschken: “I think I would like to be a physician. That was my plan initially, but it did not work out that way for me.”

Is there a scientist that you would like to have lunch with?

Joachim Büschken: “There is this German mathematician, Peter Scholze, who won the Fields medal at age 30. I tried to read up on his work but could not understand it. He created a whole new field of mathematics, and I would love to ask him ‘why?’, ‘why are you working on this?’, ‘where are you going with this?’”

Both: They have not yet met David Blei (a leading researcher in topic modeling and probabilistic models), despite their work on topic models, but would potentially like to have lunch with him.