What is MMM contribution share?

MMM contribution share is a channel's slice of the media-driven pie: its estimated contribution divided by the summed contribution of all channels, as a percentage. Where the euro contribution answers "how much margin did this channel generate", the share answers "of what marketing generated, how much came from here" — and the second question is the one the model answers most reliably.

Formula

share (channel) = channel contribution ÷ Σ all channel contributions × 100

Worked example

The model estimates search generated €260k, social €180k and display €60k — €500k of media-driven margin. The shares are 52%, 36% and 12%. If a measurement issue scaled every absolute figure down by a fifth, the shares would not move: whatever biases the response variable's level tends to hit every channel's estimate alike, which is why budget-split decisions should lean on this column.

How Saldo Metrics computes it

Each MMM run stores contribution_pct per channel in fact_mmm_channel_contribution, computed from the same fitted decomposition as the euro contribution: the channel's estimated margin over the total media-driven margin of the run (the baseline is outside the pie). v_mmm_channel_contribution serves the latest run per method. Because the ridge method's optional channel priors are themselves expressed as expected shares, this is also the quantity the priors anchor when spend histories are too correlated to separate cleanly.

Why it matters

Most MMM decisions are split decisions — how to divide next quarter's budget — and shares are exactly the currency of that decision. They are also the model's most robust output: absolute contributions inherit every scale problem in the response (cost coverage gaps, untracked revenue), while shares largely cancel those out.

Common mistakes

• Mapping shares straight onto budget. The share reports average effectiveness over the fitted period; the budget question is marginal. A high-share channel deep into saturation deserves less of the next euro than its share suggests — marginal ROAS is the column for that.
• Reading share changes between runs as trend. Different fit windows and refreshed data move estimates; compare runs qualitatively, not decimal by decimal.
• Forgetting the baseline is excluded. Shares split the media-driven margin only. A channel with a 50% share of a small media pie may matter less than the baseline's movements.

Where you see this in the app

Marketing → MMM, as the share column of the channel table.