mROAS

[AnantChandra]

Hi all!

Is there an out-of-the-box way to calculate mROAS, given that I've built a PyMC MMM?
Given a certain spend amount for a specific channel, I want to know the ROAS for any increment.

Read this paper -- https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/46001.pdf
But implementing custom logic seems non-trivial

[AnantChandra]

Went with this for now:

fig = mmm.plot_direct_contribution_curves(show_fit=True, xlim_max=1.5)

# Customize the axes labels since your target is orders
for ax in fig.axes:
    ax.set_xlabel('Daily Spend ($)')
    ax.set_ylabel('Total Daily Orders')
    ax.set_title(ax.get_title().replace('contribution', 'Total Orders'))

plt.show()

X_test_means = X_test[mmm.channel_columns].mean()

print("Mean of X_test for each channel:")
for channel, mean_value in X_test_means.items():
    print(f"{channel}: {mean_value:.2f}")
import numpy as np

marginal_cost_per_order = {}

for i, ax in enumerate(fig.axes):
   channel_name = mmm.channel_columns[i]
   mean_spend = X_test_means[channel_name]
   
   curve_line = ax.get_lines()[-1]
   x_data = curve_line.get_xdata()
   y_data = curve_line.get_ydata()

   marginal_orders = np.gradient(y_data, x_data)
   marginal_orders_at_mean = np.interp(mean_spend, x_data, marginal_orders)
   marginal_cost = 1 / marginal_orders_at_mean
   marginal_cost_per_order[channel_name] = marginal_cost
   
   print(f"{channel_name}: {marginal_cost:.2f} marginal cost per order")