context_var_ref Parameter Not Functional and Baseline Interpretation Gap

[prathyusha2510]

Problem Description:

I am trying to model binary factor variables (prime_promo, BFCM_promo) and facing a major challenge in interpreting the decomposition results due to the hidden nature of the baseline effect (the "non-promo" state).
I attempted to use the context_var_ref parameter, which appears to be designed for this exact purpose, but found it has no impact on the model output.

Input Code:

InputCollect <- robyn_inputs(
  dt_input = data,
  dt_holidays = dt_prophet_holidays,
  date_var = "date",
  dep_var = "Amazon_revenue",
  dep_var_type = "revenue",
  prophet_vars = c("trend", "season","holiday"),
  prophet_country = "US",
  context_vars = c("prime_promo","BFCM_promo"),
  context_var_ref = list(
    prime_promo = 0, # Use 0 as baseline - for non promo
    BFCM_promo = 0),
  paid_media_spends = c("amazon_daily_spend","amazon_dsp_spend"),
  paid_media_vars = c("amazon_daily_spend","amazon_dsp_spend"),
  organic_vars = c("Shopify_spends"),
  factor_vars = c( "prime_promo","BFCM_promo"),
  window_start = min(data$date),
  window_end = max(data$date),
  adstock = "weibull_pdf"
)

Observation:

1. Interpretation Gap( Reversed contribution) : For the binary variable BFCM_promo (0 = No Promo, 1 = Promo), the decomposition is inverted — it shows contribution on non-promo days (0) and zero contribution on promo days (1) but prime_promo is giving correct results.

2. Functional Failure: Running the model with context_var_ref = list(prime_promo = 0,BFCM_promo = 0) yields the exact same pareto_decomp_matrix as running the model without the context_var_ref argument entirely.

3. Missing documentation : The context_var_ref parameter is not present in the official robyn documentation.

Requested Action:

1. I am trying to model binary factors (prime_promo, BFCM_promo) but the decomposition for BFCM_promo shows reversed contributions and no effect from context_var_ref.

2. Can someone please confirm whether the context_var_ref parameter is currently functional or still under development?

3. Guidance or documentation on correctly setting the baseline for binary context variables would be very helpful.