Skip to main content
A “Model” is a statistical analysis of your marketing data — a specification of which channels, campaigns, KPI and date range Yapamar should use to answer the question “what drove our sales?”. A “Model version” is one specific fitted output produced from that model’s specification. The first version is created when you initially fit the model. Additional versions are produced automatically as new data arrives — Yapamar re-fits published models on a weekly/monthly schedule so their answers stay current with the fresh numbers.

​
Model Type

Yapamar supports 3 different kinds of analysis, each answering a slightly different question:
  • Marketing Mix — Use when your model includes a mix of digital and offline channels (TV, Radio, Outdoor, Press alongside Facebook, Google, TikTok, etc.). Measures the combined impact across all media types on your KPI. This is the safer default and works for any channel mix.
  • Digital Attribution — Use when measuring only digital channels (Facebook, Google, TikTok, Instagram, etc.). Tuned for the digital media. Offline channels (TV, Radio, Outdoor, Press) aren’t supported in this mode — if you need them, pick Marketing Mix instead.
  • Long-Term Effect — Use when you want to see beyond the immediate sales impact and measure how your advertising builds the brand over time. The Long-Term model estimates the lasting effect: how media investment strengthens brand awareness: tracked through your brand’s share of search and how that translates into baseline sales over the following months, long after a campaign has ended.
You choose the model type when creating a model. It can be changed later by editing the model, which re-fits the specification from scratch using the new priors and any updated channel selection. A Long-Term model isn’t built from scratch — it’s created on top of an existing Marketing Mix model. When you create one, you simply pick which Marketing Mix model it should be based on; the KPI, media channels, and campaign selection are taken over automatically, and the Marketing Mix results serve as its starting point. This means the two models always tell one consistent story: short-term effects from the Marketing Mix model, long-term effects layered on top.