Dylan Poulsen Consulting
Thoughtful statistical consulting for clear, defensible decisions
Thoughtful statistical consulting for clear, defensible decisions
I help teams decide what worked, what to do next, and how much uncertainty remains, then turn the analysis into clear reports and reusable workflows that make future work easier.
As featured in
Consulting shaped by research, teaching, and practice
I bring a research background in data science and mathematics to consulting work that needs both rigor and judgment.
Evaluate whether a policy, intervention, treatment, program, or operational change plausibly caused the outcome you care about.
Turn scattered spreadsheets, research files, or operational exports into clean datasets, documented analysis, and visualizations people trust.
Replace manual recurring reports with Quarto, R, Python, dashboards, and automation that reduce drudgery and can be rerun without redoing the work by hand.
Build or review models with documented assumptions, uncertainty, limitations, and language nontechnical colleagues can act on.
I work best on decision problems where the analysis must be rigorous, explainable, and reusable. The tools vary by project; the core focus is helping people act wisely under uncertainty.
Bayesian and causal reasoning for intervention questions, observational studies, experiments, forecasting, and decision support.
Quarto reports, Shiny apps, dashboards, and documented routines that replace fragile manual analysis with calmer, repeatable systems.
Careful reviews of models, datasets, and reports for interpretability, privacy risk, uncertainty, and whether the analysis truly supports the decision.
Clarify whether your data can support a causal claim, identify the strongest analysis strategy, and deliver a plain-language decision memo.
Convert a recurring manual report into a reusable Quarto, R, Python, or dashboard workflow with documented assumptions and update steps.
Review an existing model, dataset, or analysis workflow for uncertainty, interpretability, privacy, and whether the conclusions are well supported.
I lead applied data science work in both consulting and academic settings, with a focus on reproducible systems that support real decisions under uncertainty.
As Director of the Data Science Innovation Lab at Washington College, I also supervise project-based analytics work that keeps my consulting practice grounded in real data, real people, and useful delivery.
Bayesian workflow for client work
This is the same process I use in client work, from first conversation to final handoff.
Problem: Assess RNA quality quickly using instrument-derived data.
Approach: I built a machine learning workflow with Thermo Fisher collaborators to estimate RNA quality from instrument-derived features.
Deliverable: The project produced a model-backed analysis and conference poster that translated technical performance into practical lab guidance.
Problem: Answer a viral internet question posed by chef Kenji López-Alt: how deep should you cut into an onion to make the pieces as uniform as possible?
Approach: I built a mathematical model to determine the statistical distribution of the onion piece sizes for any given cut depth, then found the optimal cut depth to minimize the variance of the sizes.
Deliverable: I wrote about the model in a fun, accessible, but rigorous way. The work made mathematical modeling accessible to broad audiences and was featured by The New York Times, The Pudding, and Daily Mail.
Problem: Data practice changes quickly, so undergraduate training can lag behind real-world analytical work.
Approach: I help lead Washington College’s Data Science curriculum, teach GitHub for collaboration and reproducible workflows, and teach Bayesian causal inference as an underused but essential framework.
Deliverable: Students leave with modern tools, reproducible teamwork habits, and portfolio-ready analyses they can maintain and adapt as methods evolve.
I bring careful methods, practical communication, and transparent uncertainty to every consulting project.