Now
What I'm working on, thinking about, and building toward.
Last updated: August 2026
What I'm working on, thinking about, and building toward.
Last updated: August 2026
This page is inspired by Derek Sivers' now page concept, a snapshot of what's actually occupying my attention rather than a static bio.
I'm currently focused on expanding the content valuation infrastructure, moving from title-level ROI analysis to portfolio-level strategy. The question I'm most interested in: how do you optimize a content library when the value of any individual title depends on what else is in the library? That's a harder problem than it looks, and it's one of the genuinely interesting optimization challenges in the streaming space right now.
Three pieces went up on Data Science Rabbit Hole in August: Why the OpenAI-Influencer Backlash Was Completely Predictable, XGBoost Does Not Care That the Same Customer Appears 52 Times, and Correlation Is Lossy Compression. That spread is roughly the publication's whole thesis: the culture around this technology, the methodological traps practitioners walk into, and the statistics underneath both.
My Data Science for Decision Makers column on All Things Insights continues. The August installment, The Modern Balanced Scorecard: Why It Still Matters, argues that the scorecard was never a reporting format. It was an operating system for connecting metrics to each other by cause and effect, and organizations that reduced it to a dashboard threw away the only part that mattered.
I'm off from the University of Oklahoma until Spring 2027, when Marketing & Media Analytics runs again. A semester away from a syllabus is a good way to find out which parts of it you actually believe.
In late September I'm giving a webinar for the National Merit Scholarship Corporation called How to Disagree with Data. There is something fitting about being asked to teach productive disagreement by the organization that named me a Scholar in 1988.
Then Denver, October 5–7, where TMRE and Content Marketing World are being held together for the first time. I'm speaking on both sides of it. The premise of putting insights people and marketing people in one building is that the handoff between them is where the value leaks out, which happens to be the thing I have spent a decade writing about.
Monday afternoon I'm on one of the two teams in The Future of Insights Throwdown, a general session that puts three insights leaders across from three others and lets them argue. Tuesday morning, on the Marketing Analytics, Data & ROI track at Content Marketing World: You Can't Measure Success by Success, on why a campaign that worked is not evidence that anyone decided well. Timing, luck, budget, and a shifting audience will all produce a rising line, and none of them are reasoning. Tuesday afternoon at TMRE: The Missing Layer: Why Great Analytics Still Fails to Drive Decisions, which builds a decision-layer framework out of Warren Powell's work at Princeton and gets concrete about matching the type of analysis to the type of decision.
I'm increasingly interested in the question of organizational data maturity, not just technical capability but the decision-making culture that determines whether technical capability translates into business value. Most organizations that think they have a data science problem actually have a decision science problem, and the fix isn't more sophisticated models.
The other thing occupying me is the growing anti-AI sentiment, which seems strongest among people considerably younger than me. It would be easy to write off as reflexive, and I don't think it is. We have had roughly four decades in which technological advancement was treated as self-evidently good and the only respectable question was how fast. That consensus is gone. Nothing has replaced it.
Which leaves a question I have not resolved: how do I stay good at this job, honest about my own ethics, and aligned with what society is now asking of the people who build these systems? Those three used to point in the same direction. They no longer do so automatically. I would rather sit in that discomfort than resolve it early, and I'm most suspicious of the people who have already resolved it in either direction.
Two of my projects are now real enough to pip install. CorrSleuth runs several complementary association measures on the same pair of variables and tells you in plain English what kind of relationship you actually have: nonlinear, non-monotonic, or driven by four points you should probably go look at. Foreledger is an append-only archive for forecasts, so that when a forecast turns out to be wrong you can still find out what it originally said. Most organizations overwrite their predictions with their results and then wonder why nobody trusts the model.
Not all of it is data science. I have not seen the Odyssey film and am waiting for streaming, but I have read the poem twice and it is one of my favorites, and the sheer volume of noise around the movie shook loose an EP's worth of material. Eight tracks, called Odyssey, released under my hip-hop project The Indefatigable Horde. Homer has survived worse. I spend most of my writing life arguing that hype is a bad guide to decisions. It turns out to be a fine guide to writing songs.
I'm interested in problems at a larger scope: enterprise-level data strategy, AI governance, and the organizational design questions that determine whether advanced analytics creates durable competitive advantage or expensive overhead. If you're working on those problems and want to think through them together, I'm reachable by email.