Skip to content

Releases

MMM Engine v0.2.1

MMM Engine v0.2.1 Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 25 Oct 21:17
v0.2.1
49cebb0
This commit was signed with the committer’s verified signature.
claude Claude
SSH Key Fingerprint: 32dP45eSMmVSt/G/CGvcxl/P+MO3Nwj9xeTh/GSA2wc
Verified
Learn about vigilant mode.

Added

  • mmm fit refuses channels the data cannot identify: constant spend, or two channels whose weekly spend correlates at r >= 0.98 (--max-correlation). Warns from r >= 0.90; warnings also go to fit.warnings in the model JSON.
  • macOS Apple Silicon (arm64) release binary; macOS build in CI.

Install

curl -LO https://github.com/yablokolabs/mmm-engine/releases/download/v0.2.1/mmm-v0.2.1-linux-x86_64.tar.gz
tar -xzf mmm-v0.2.1-linux-x86_64.tar.gz
./mmm-v0.2.1-linux-x86_64/mmm help

Linux binaries need glibc 2.38+ (Ubuntu 24.04+, Debian 13+, Fedora 39+). macOS: Apple Silicon. Older systems: use Docker.

Docker: docker run --rm -v "$PWD:/work" ghcr.io/yablokolabs/mmm-engine:0.2.1 help

Assets 8
Loading

MMM Engine v0.2.0

MMM Engine v0.2.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 24 Oct 20:53
v0.2.0
f9b8542
This commit was signed with the committer’s verified signature.
claude Claude
SSH Key Fingerprint: 32dP45eSMmVSt/G/CGvcxl/P+MO3Nwj9xeTh/GSA2wc
Verified
Learn about vigilant mode.

Added

  • mmm fit: learns adstock decay, Hill half-saturation and slope per channel (coordinate grid search), with trend, seasonality and non-negative channel betas by least squares. Scores the last --holdout weeks on a model that never saw them.
  • mmm optimize: budget reallocation by greedy marginal return plus pairwise exchange, each channel within --bound of current spend. Reports lift and marginal ROI.
  • CSV input with line-numbered errors.
  • examples/synthetic: 3 years of data with known answers, run.sh, grade.py.
  • Test suite (mmm_tests): known-truth recovery, determinism, optimizer vs brute force, decision quality on the true curves.
  • CI on every push and pull request; releases (tag push or the "Release" workflow's Run workflow button) ship Linux x86-64 and arm64 binaries and a Docker image on GHCR.
  • MMM_NATIVE CMake option (default ON); release builds turn it off for portable binaries.

Changed

  • mmm forecast future weeks include media driven by a spend plan (--plan), or current spend by default. Previously only trend and seasonality.

Removed

  • The placeholder budget optimizer that drifted toward an equal split.

Install

curl -LO https://github.com/yablokolabs/mmm-engine/releases/download/v0.2.0/mmm-v0.2.0-linux-x86_64.tar.gz
tar -xzf mmm-v0.2.0-linux-x86_64.tar.gz
./mmm-v0.2.0-linux-x86_64/mmm help

Needs glibc 2.38+ (Ubuntu 24.04+, Debian 13+, Fedora 39+). Older systems: use Docker.

Docker: docker run --rm -v "$PWD:/work" ghcr.io/yablokolabs/mmm-engine:0.2.0 help

Loading