Optimizations

Purpose of this page

The Optimization View is designed to help you make data-driven spend decisions at the campaign and sub-channel level. Using predictions from the Magpie AI Model, this page provides spend recommendations based on profitability curves.

It’s built to support weekly optimization, helping you reallocate budgets intelligently and forecast future outcomes.

How it works

Prediction Logic

The Magpie Model forecasts profitability for each campaign or sub-channel by estimating:

• The expected profit curve over increasing spend levels

• Where your current weekly spend sits on that curve

• The suggested spend level that would improve your profit

• Curves are only estimated for campaigns with enough data points. Smaller campaings are not shown in this overview. But can be run manually, reach out if you would be interested.

All recommendations and charts are based on weekly data.

📌 You cannot change the time granularity on this page. Daily or monthly views are not supported here.

Filters and Dimension Control

You can:

• Filter by Channel, Sub-channel, or Campaign

• Optimizations are only shown for sub channel or campaign level. On channel level results are nog accurate enough, due to cluttered strategies. (e.g. the dynamics behind Branded search are totally different from Pmax for Google Ads). Toggle the view by using the "show optimization for" button at the top of the chart

Date Logic

• You cannot adjust time granularity (only week-level is used). No "day" or "month" breakdowns are available on this view

• You cannot select a period. Results are automatically generated based on the marketing spend of last 10 -15 weeks.

Use filters to target the specific part of your media mix you'd like to optimize.

Spend Recommendation Logic

For each campaign or sub-channel, the model estimates:

• Current spend level (based on the most recent full week)

• Optimal spend level (based on the model's prediction)

Content

Optimization Table

At the top of the page, you’ll find a results table showing:

• Last week's actuals per campaign or sub-channel

• The model's suggested optimal spend

• The estimated uplift in profit if the suggestion is applied

• Additional columns may include:

  • Last week's margin, cost, and profit

  • Predicted values at the optimal spend level

  • Potential overspend or underspend gap

Curves

Profit Curve

This graph shows:

• The predicted profit at different spend levels

• A yellow dot indicating your actual spend from last week

• A curve that helps you understand whether you're under, over, or optimally spending

Use this to quickly assess if there's room to gain more profit or if you're entering diminishing returns territory.

Diminishing Returns Curve

This second graph focuses on:

• The margin (revenue – cost) rather than profit

• How expected margin changes with increasing spend

• Helps identify how efficient each euro/dollar becomes as you increase investment

It’s ideal for understanding scale potential vs efficiency loss.

Curve Fit Quality

This graph shows how well the model's predictions align with actual data:

• Includes normalized data points from the last 10–15 weeks

• Helps you understand the confidence and stability of the model. The closer the dots are to the green line, the more accurate the estimations.

• Visualizes the datapoints used for training. A broad selection of dots indicates a variety of training data. In this example, the training data is between 8K and 16K spend per week. Predictions higher than 16K and lower than 8K are less reliable.

• A strong fit indicates a reliable optimization suggestion, while weak or noisy fits should be interpreted more cautiously

This table gives you a quick overview of where optimizations are possible, and where spend is already close to optimal.

Use this page to:

• Adjust campaign-level budgets for maximum weekly profit

• Spot over- or under-invested campaigns/sub-channels

• Identify where additional spend will have limited impact

• Evaluate the reliability of optimization suggestions based on curve fit quality

• Act on weekly optimization opportunities with clear data-backed recommendations

Want to understand how the profit and margin curves are built?
See the Methodology – Curves section for a full explanation.