Product updates, engineering deep dives, and marketing science perspectives from the SIMBA team.
What our voice-agent evaluations taught us about speed, corrections and user control, and how those lessons inform the way we want people to work with Simba.
Regional sales do not make national media regional. When a geo MMM adds causal information, why population-apportioned spend fails, and a quick diagnostic.
A Bass model needs p, q and M before a new product has sold anything. Learn them from previous launches, then update as the first weeks of sales arrive.
Budget optimisation has two sensible modes. Without a trusted value per acquisition, hold spend fixed and maximise acquisitions. With one, let the budget move.
R-hat 1.01 came from a March 2019 paper, reached PyMC in a single 2022 pull request, and never became a universal law. Here is the actual timeline.
Compress a reference posterior into Gaussian likelihood messages, solve candidate priors analytically, and predict NUTS ESS before paying for the refit.
Delayed adstock trades pure decay for build, peak and decay. Keep the half-life parameterisation, and the two questions it lets your team argue about.
Why channel removal effects overlap in a multiplicative MMM, how Shapley and Aumann-Shapley allocate that overlap, and what benchmark tests tell us about the trade-offs.
Standard MMMs score brand ROI near zero. A two-stage workflow, time-varying baseline plus Bayesian VARX, now ships in PyMC-Marketing 1.0. Here's how it works.
PyMC-Marketing 1.0 ships a stable API, multidimensional MMM by default, and roughly 1.5x faster sampling. What changed, what breaks, and how to upgrade.
The standard geometric adstock rate means nothing in a marketing meeting. Reparameterize it in half-life units with a two-line change everyone understands.
A practical comparison of PyMC-Marketing and Google Meridian for teams choosing a Bayesian marketing mix modeling workflow.
When PyMC-Marketing notebooks are enough, when they become hard to operationalize, and where a UI like SIMBA helps business teams use Bayesian MMM.
A plain-English guide to the Bayesian MMM concepts behind channel ROI, response curves, uncertainty, adstock, saturation, priors, and lift-test calibration.
A practical checklist for building the weekly dataset your MMM needs: media spend, sales, controls, promotions, external factors, and clean naming.
MMM does not fail because of bad math. It fails because getting clean, consistent data from ten platforms into one place is still a manual nightmare.
Build Bayesian MMMs, measure channel ROI, optimize budgets, and run scenario forecasts through natural language in Claude, Cursor, or any MCP client.