Pulsar Analytics
Book

SERVICES

Three AI-powered engines. Start with comparison.

Each engine is an AI-powered product. The language model does not invent the numbers — it writes from a context stack built from ten years of marketing-analytics practice.

The system is built that way: journeys, models, scoring, draft, professional review, your PDF.

The diagnostic is how the starting engine is chosen. Path credit and mix-model numbers are not blended into one fake truth.

01

The starting point

Attribution Model Comparison Engine

The engine connects to the GA4 property, builds customer journeys from the BigQuery export, and runs several models at once — including Markov chains and Shapley values — on AWS. An evaluator scores them against simple checks. A draft PDF is written automatically. Every report is approved by a marketing professional with extensive experience in marketing analytics before it is sent.

  • Journeys from your GA4 BigQuery export — not the Data API
  • Markov chains, Shapley values, and four simple path rules, side by side
  • Scored against stability checks, with a language model that cannot invent numbers
  • Weekly ranked PDF, approved by a marketing professional before send.
Read the full page

02

When you have the history

Media-Mix Model as a Service

The service ingests GA4 and weekly spend, runs Meridian or PyMC models on AWS, and sends a budget-reallocation PDF. The agent retrains on a weekly cadence, checks diagnostics, writes recommendations in plain language, and emails the PDF after professional review. This needs a clean spend history. If it is not there, that is said up front.

  • GA4 plus spend in, Meridian or PyMC on AWS
  • Weekly retraining with diagnostic checks
  • Budget-reallocation PDF, not a live dashboard
  • Path credit and mix models stay separate numbers — they are not blended
Read the full page

03

From credit to a plan

Attribution Budget Optimizer

The optimizer takes GA4-attributed return, adds the min/max channel caps, and solves the allocation with PuLP. The agent pulls fresh data on a nightly cadence, re-solves, writes scenario notes in plain language, and ships a media-plan PDF on schedule. Labeled as a scenario — not as truth.

  • Constrained allocation with PuLP on your caps and brand rules
  • Nightly refresh from GA4-attributed return
  • Scenario notes in the PDF, clearly labeled
  • Sits on top of the comparison engine — it does not replace it
Read the full page
Book a diagnosticWhat’s inside the weekly report