Attribution
Model, lookback, and the metric itself
Attribution in CJA distributes credit for a conversion across the touchpoints that preceded it. The exam expects you to know the models (Last Touch, First Touch, Linear, Participation, Same Touch, U-shaped, J-shaped, Inverse J, Time Decay, Custom, and Algorithmic) plus the two settings that change results as much as the model does: the lookback window and the metric being attributed.
Attribution is configured per metric in the data view, so your organization's default lives in one place and every Workspace column inherits it. Analysts can override it on a column in a table or Attribution panel, and that override is the correct way to test an alternative model without retraining everyone.
Traps. Participation gives full credit to every touchpoint, so totals can far exceed conversion counts; Same Touch credits every touchpoint as though it were the last, which is not the same thing. Only partitionable metrics split fractional credit, so a person or order count may not divide cleanly. The lookback window applies to the touchpoint, not the conversion, so shortening it can erase upper-funnel channels from reporting. Algorithmic attribution needs enough conversion volume to be stable. Changing the data view default rewrites history for every user, retrospectively.
Rule of thumb: compare Last Touch against Algorithmic side by side before recommending budget shifts, and always state the lookback window alongside the model. Worked example: a 30-day lookback omits the first touch of a 45-day enterprise sales cycle; extending the window materially raises the display channel's credited revenue.