optimizer.BudgetOptimizer

[asos-jenniferlim]

Hi

I am trying to increase my budget for the next fiscal year from 40m to 45m (all media channels except crm_push and crm_email), but after adding current budget with the delta optimisation - the sum is the same?

budget_optimizer = optimizer.BudgetOptimizer(mmm)

spend_constraints_lower = [0.3] * num_media_channels
spend_constraints_upper = [0.3] * num_media_channels

# Budget for CRM not changed so setting these to 0
crm_email_index = media_channel_list.index('crm_email')
crm_push_index = media_channel_list.index('crm_push')
for arr in (spend_constraints_lower, spend_constraints_upper):
    arr[crm_email_index] = 0.0
    arr[crm_push_index] = 0.0

budget = 45_500_000
optimization_results = budget_optimizer.optimize(
    fixed_budget=True,
    budget=budget,
    start_date='2024-07-01',
    end_date='2025-06-23',
    spend_constraint_lower=spend_constraints_lower,
    spend_constraint_upper=spend_constraints_upper
)
optimization_results.output_optimization_summary(optimisation_output_filename, output_dir)
optimized_spend = optimization_results._get_delta_data('spend')
optimized_spend_indexed = optimized_spend.set_index('channel')

# Joined back to previous budget
sm = analyzer.Analyzer(mmm).summary_metrics(
    confidence_level=0.8,
    selected_times=list(mmm.input_data.time[-52:].data),
)
sm_df = pd.DataFrame(
    data={
        "channel": sm.channel[:-1],
        "previous_spend": sm.spend[:-1]
    }
)

df = sm_df.join(optimized_spend_indexed, on="channel", how='left')
df.loc[:, 'optimized_spend'] = df['previous_spend'] + df['spend']
df

So df['optimized_spend'].sum() is the same as df['previous_spend'].sum()

If this is not the right approach, how do I get the new optimised budget?

[gousesb]

Hello @asos-jenniferlim ,

Thank you for contacting us!

To correctly evaluate the optimization, instead of using delta data, the recommended approach is to directly use the optimized spend from the optimization results and compare it with historical spend from the Analyzer. The optimized spend will reflect the new budget allocation, while the historical spend will reflect the baseline budget. When summed up, the historical spend should show a total baseline of 40M, and the optimized spend should show the new budget of 45.5M.

For instance, let’s say the historical spend is split as follows: TV = 15M, Digital Display = 10M, Search = 15M, CRM Email = 0, CRM Push = 0. The total baseline here is 40M. After running the optimization with a fixed budget of 45.5M, the optimized allocation might shift to: TV = 16.5M, Digital Display = 12M, Search = 17M, CRM Email = 0, CRM Push = 0. The new total equals 45.5M. This way instead of adding baseline and delta together, you can directly compare historical versus optimized budgets. A channel-level comparison shows how the extra 5.5M is distributed across TV, Digital, Search, and Social, while CRM channels remain unchanged.

optimized_spend = optimization_results.optimized_data.spend

historical_spend = analyzer.Analyzer(mmm).get_historical_spend()

Feel free to reach out for any further queries or feedback regarding Meridian.

Google Meridian Support Team