reward_model

reward_model performs attribution from aggregated signals using a movement/coherence-based reward baseline. It uses the same tabular D_variables structure as auto_mmm.

At a high level, auto_mmm estimates predictive contribution with the automatic MMM engine, while reward_model estimates contribution from directional movement coherence between signals and the target.

Input structure

Data should contain:

• a timestamp column
• one target column, passed separately through the target parameter
• one or more signal columns used as attribution inputs

D_variables should be a table/data frame with one row per non-target signal variable and these columns:

COLUMN	DESCRIPTION
variable	Name of the signal column in Data.
channel	Final attribution channel associated with the signal. Multiple variables can map to the same channel.
measure	Measure type for the signal, such as impressions, clicks, cost, direct_searches, platform_signal, or another supported measure label.

The old D_measures parameter is no longer used. Measure information must be provided inside D_variables.

Example D_variables structure:

variable              channel       measure
direct_searches       direct        direct_searches
facebook_impressions  facebook_ads  impressions
facebook_cost         facebook_ads  cost
google_impressions    google_ads    impressions
google_clicks         google_ads    clicks
google_cost           google_ads    cost

Parameters

PARAMETER	TYPE	DEFAULT	DESCRIPTION
Data	data.frame		Aggregated dataset with timestamp, target, and input signal columns.
D_variables	data.frame / table		Table with variable, channel, and measure columns.
target	str	"conversions"	Name of the target column in Data.
max_p	int	12	Maximum lag length considered when the model evaluates delayed effects.
seed	int	1234567	Random seed for reproducible results.
verbose	int / bool	1	Controls runtime logging.
server	str / list[str]	app.channelattribution.io	Server endpoint used by the hosted computation service.
password	str	NULL / None	Authentication token for the hosted service.
return_diagnostics	bool	FALSE / False	If enabled, returns attribution together with diagnostic objects for advanced inspection.

Scoring summary

For each channel and lag, reward_model evaluates directional coherence between the lagged signal and the target. The lag score uses the number of non-zero observations used for the lag as a reliability factor:

score_lag = n0 / (n0 + 10) * coherence_rate

where n0 is the number of non-zero observations used to evaluate the current lag. The selected channel score is then converted into a score_multiplier through the shared median/blend scoring logic.

Output

If return_diagnostics = FALSE, the function returns an attribution data frame.

If return_diagnostics = TRUE, the function returns a list/dict containing:

OUTPUT	TYPE	DESCRIPTION
attribution	data.frame	Channel attribution by time period.
base_raw	data.frame / object	Raw baseline attribution output before final adjustments.
score_multiplier	dict / object	Score multiplier information used during attribution assembly.
applied_score_multiplier	dict / object	Effective multiplier applied to the returned attribution.
individual_scores	dict / object	Channel-level movement/coherence scores selected from the lag evaluation.
mode	str	Execution mode metadata.
diagnostics	list / dict	Additional diagnostics intended for advanced inspection.