Skip to main content

MixShift Intelligence

MixShift Intelligence explains why a number moved, not just what it is. Instead of a dashboard you have to interpret, it returns a finished result: the headline change, a bridge whose pieces sum exactly to that change, the caveats the engine raised on its own, and a stamped record of the version it was computed with. It is open to every account. See Get access for what that costs.

How the pieces fit together

Intelligence is organized in a few layers, and it helps to know their names, because you will see all of them:

  • MixShift Intelligence is the product: the umbrella you sign in to, and the credential scope you grant to your own tools.
  • Underneath it are services, the broad jobs Intelligence does. Attribution is live today. Forecasting serves a readiness check, the forecast baseline, and a report-month tracker; see Forecasting.
  • Each service is made up of requests: the specific, named things you can ask for, each metered on its own. Attribution's requests today are the Operational Bridge, the Advertising Bridge, the Combined Read, Lost Sales, and the Monthly Read (the full list is in What Intelligence answers).
  • Underneath all of it is the method, how the engine actually gets from raw data to an answer. Attribution runs on HCAM, the Hierarchical Causal Attribution Model, with two views: through which levers performance moved, and where in the business it moved. See Reading a bridge for how to read one, and the whitepaper, RAG-Enabled Hierarchical Causal Attribution Model, for the method in full.

Every service reads the same brand memory your team already maintains, the brand context and change log on the Builder Platform. A report built on a bridge can explain a move with what your team already recorded there, a launch, a promotion, a stockout, while the engine's own caveats come from the data. See Work with brand memory.

Attribution, today

Attribution answers five requests: the Operational Bridge, the Advertising Bridge, the Combined Read, Lost Sales, and the Monthly Read. Every result is a JSON envelope with a headline, a footing-checked bridge, a list of caveats, a limitations list, and a stamp of the engine and evidence versions it ran on. Add evidence to a request and it also returns "What we know": measured facts and candidate explanations, each with its own evidence. See What Intelligence answers.

Forecasting

Forecasting is a service under MixShift Intelligence, on the same catalog, gate, and metering as Attribution. It answers a readiness check (whether a merchant has the history to support a forecast) and serves the forecast baseline and the report-month tracker as data. The baseline is the published forecast, or a revenue baseline computed from your warehouse data and labelled as computed when nothing is published; when there is not enough history, the answer says so. Your team publishes and corrects the forecast in the MixShift forecasting app, and in Claude the Forecast Max skill explains it in plain language with its measured accuracy. See Forecasting and Get a forecast with Forecast Max.

The doors

Getting started

How to

Concept

Reference

FAQ

Still have questions?
Ask Shifty

Shifty is the MixShift docs assistant. Pick a question or open the chat for anything else.

Back to Knowledge Base