robyn_allocator: Budget Allocator

In Robyn: Semi-Automated Marketing Mix Modeling (MMM) from Meta Marketing Science

Budget Allocator

Description

robyn_allocator() function returns a new split of media variable spends that maximizes the total media response.

Usage

robyn_allocator(
  robyn_object = NULL,
  select_build = 0,
  InputCollect = NULL,
  OutputCollect = NULL,
  select_model = NULL,
  json_file = NULL,
  scenario = "max_response",
  total_budget = NULL,
  target_value = NULL,
  date_range = "all",
  channel_constr_low = NULL,
  channel_constr_up = NULL,
  channel_constr_multiplier = 3,
  optim_algo = "SLSQP_AUGLAG",
  maxeval = 1e+05,
  constr_mode = "eq",
  plots = TRUE,
  plot_folder = NULL,
  plot_folder_sub = NULL,
  export = TRUE,
  quiet = FALSE,
  ui = FALSE,
  ...
)

## S3 method for class 'robyn_allocator'
print(x, ...)

## S3 method for class 'robyn_allocator'
plot(x, ...)

Arguments

robyn_object	Character or List. Path of the Robyn.RDS object that contains all previous modeling information or the imported list.
select_build	Integer. Default to the latest model build. select_build = 0 selects the initial model. select_build = 1 selects the first refresh model.
InputCollect	List. Contains all input parameters for the model. Required when robyn_object is not provided.
OutputCollect	List. Containing all model result. Required when robyn_object is not provided.
select_model	Character. A model SolID. When robyn_object is provided, select_model defaults to the already selected SolID. When robyn_object is not provided, select_model must be provided with InputCollect and OutputCollect, and must be one of OutputCollect$allSolutions.
json_file	Character. JSON file to import previously exported inputs or recreate a model. To generate this file, use robyn_write(). If you didn't export your data in the json file as "raw_data", dt_input must be provided; dt_holidays input is optional.
scenario	Character. Accepted options are: "max_response", "target_efficiency". Scenario "max_response" answers the question: "What's the potential revenue/conversions lift with the same (or custom) spend level in date_range and what is the allocation and expected response mix?" Scenario "target_efficiency" optimizes ROAS or CPA and answers the question: "What's the potential revenue/conversions lift and spend levels based on a target_value for CPA/ROAS and what is the allocation and expected response mix?" Deprecated scenario: "max_response_expected_spend".
total_budget	Numeric. Total marketing budget for all paid channels for the period in date_range.
target_value	Numeric. When using the scenario "target_efficiency", target_value is the desired ROAS or CPA with no upper spend limit. Default is set to 80% of initial ROAS or 120% of initial CPA, when "target_value = NULL".
date_range	Character. Date(s) to apply adstocked transformations and pick mean spends per channel. Set one of: "all", "last", or "last_n" (where n is the last N dates available), date (i.e. "2022-03-27"), or date range (i.e. c("2022-01-01", "2022-12-31")). Default to "all".
channel_constr_low, channel_constr_up	Numeric vectors. The lower and upper bounds for each paid media variable when maximizing total media response. For example, channel_constr_low = 0.7 means minimum spend of the variable is 70 average, using non-zero spend values, within date_min and date_max date range. Both constrains must be length 1 (same for all values) OR same length and order as paid_media_spends. It's not recommended to 'exaggerate' upper bounds, especially if the new level is way higher than historical level. Lower bound must be >=0.01, and upper bound should be < 5.
channel_constr_multiplier	Numeric. Default to 3. For example, if channel_constr_low and channel_constr_up are 0.8 to 1.2, the range is 0.4. The allocator will also show the optimum solution for a larger constraint range of 0.4 x 3 = 1.2, or 0.4 to 1.6, to show the optimization potential to support allocation interpretation and decision.
optim_algo	Character. Default to "SLSQP_AUGLAG", short for "Sequential Least-Squares Quadratic Programming" and "Augmented Lagrangian". Alternatively, ""MMA_AUGLAG", short for "Methods of Moving Asymptotes". More details see the documentation of NLopt here.
maxeval	Integer. The maximum iteration of the global optimization algorithm. Defaults to 100000.
constr_mode	Character. Options are "eq" or "ineq", indicating constraints with equality or inequality.
plots	Boolean. Generate plots?
plot_folder	Character. Path for saving plots and files. Default to robyn_object and saves plot in the same directory as robyn_object.
plot_folder_sub	Character. Sub path for saving plots. Will overwrite the default path with timestamp or, for refresh and allocator, simply overwrite files.
export	Boolean. Export outcomes into local files?
quiet	Boolean. Keep messages off?
ui	Boolean. Save additional outputs for UI usage. List outcome.
...	Additional parameters passed to robyn_outputs().
x	robyn_allocator() output.

Value

A list object containing allocator result.

List. Contains optimized allocation results and plots.

Examples

## Not run: 
# Having InputCollect and OutputCollect results
AllocatorCollect <- robyn_allocator(
  InputCollect = InputCollect,
  OutputCollect = OutputCollect,
  select_model = "1_2_3",
  scenario = "max_response",
  channel_constr_low = 0.7,
  channel_constr_up = c(1.2, 1.5, 1.5, 1.5, 1.5),
  channel_constr_multiplier = 4,
  date_range = "last_26",
  export = FALSE
)
# Print a summary
print(AllocatorCollect)
# Plot the allocator one-pager
plot(AllocatorCollect)

## End(Not run)

Robyn documentation built on June 27, 2024, 9:06 a.m.